Home Docs Claude Developer Platform 30 Oct 2026

One change · api

optimizing-for-cost-and-intelligence changedabout-claude/models/optimizing-for-cost-and-intelligence

Nearest release: v2.1.315, published under an hour before this site recorded the change. Shown because the two are within 24 hours of each other. Nothing here says the release caused the edit.

Recorded here 30 Oct 2026 21:26 UTC

Lines+101added

Lines−101removed

From line 57 where the diff opens

First seen 14 Aug 2026 this site's first read of the page

Recorded edits33to this page, all time

The whole hunk

from line 57, old and new numbered

lines

from line 57

5757

5858**What good looks like.**

5959

60* Over a full day of real traffic, agent loops read a median 84% of their input from the cache, and the top 10% of them, coding or not, read 94% or more[17](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

60* Over a full day of real traffic, agent loops read a median 84% of their input from the cache, and the top 10% of them, coding or not, read 94% or more[17](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

6161* Deep in a task, a well-built loop pays full price on under 1% of its input.

6262* Below about 80%, look for something breaking the cache (see [What breaks the cache](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#what-breaks-the-cache)).

6363

64Anthropic priced the DeepResearch Bench II[7](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) runs with and without the cache discount:

64Anthropic priced the DeepResearch Bench II[7](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) runs with and without the cache discount:

6565

6666<Frame>

6767 ![Chart, DeepResearch Bench II: caching cuts Fable 5.1 from $40.58 USD to $5.64 USD a task; Sonnet 5.5, $5.13 USD to $1.05 USD](https://platform.claude.com/docs/images/cost-intel/caching-repriced.svg)

from line 71

7171

7272Pricing other runs’ recorded token counts the same way, each model at its default effort, gives the average token cost per task without caching:

7373

74| Benchmark | Model | Without caching |

75| ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------- | ------------------------------- |

76| SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Opus 5.5 | 3.3 times the cost with caching |

77| SWE-bench Pro | Claude Sonnet 5.5 | 3.1 times the cost with caching |

78| Claude Managed Agents task set[25](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), whose sessions are short | Claude Haiku 5.5 | 1.7 times the cost with caching |

79| A small agent that triages bug reports ([Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens))[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Sonnet 5.5 | 7.3 times the cost with caching |

74| Benchmark | Model | Without caching |

75| ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------- | ------------------------------- |

76| SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Opus 5.5 | 3.3 times the cost with caching |

77| SWE-bench Pro | Claude Sonnet 5.5 | 3.1 times the cost with caching |

78| Claude Managed Agents task set[25](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), whose sessions are short | Claude Haiku 5.5 | 1.7 times the cost with caching |

79| A small agent that triages bug reports ([Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens))[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Sonnet 5.5 | 7.3 times the cost with caching |

8080

8181The saving varies with how many turns a task takes, because shorter loops re-read less.

8282

from line 83

8383<Note>

8484 With Claude Haiku 5.5 on the Messages API, [compact a long conversation](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#manage-the-context-lifecycle) between tasks, before its prompt passes 100,000 tokens. Past that, every token of the request costs 5 times as much, output included. Cache reads and writes count toward the 100,000 (see [Long context pricing](https://platform.claude.com/docs/en/about-claude/pricing#long-context-pricing)).

8585

86 If you use [compaction at a token threshold](https://platform.claude.com/docs/en/build-with-claude/compaction-threshold#parameters) instead, set its `trigger` a little below 100,000 minus your prompt’s largest growth between requests (the default is 150,000). In Anthropic’s test with 10,000-token turns, a trigger of 90,000 cost 70% less per request than the default[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

86 If you use [compaction at a token threshold](https://platform.claude.com/docs/en/build-with-claude/compaction-threshold#parameters) instead, set its `trigger` a little below 100,000 minus your prompt’s largest growth between requests (the default is 150,000). In Anthropic’s test with 10,000-token turns, a trigger of 90,000 cost 70% less per request than the default[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

8787</Note>

8888

8989#### Pick the cache duration

from line 92

9292

9393The 1-hour duration costs more to write (2x the input price instead of 1.25x). A miss on either duration bills the whole prefix at the write price instead of the read price.

9494

95Choose the cache duration when a conversation starts. In Anthropic’s test, the first turn added at 1 hour to a conversation cached at 5 minutes wrote the whole prompt again, once: $0.80 USD at 100,000 tokens on Claude Opus 5.5, compared with $0.02 USD of input for a turn that read the cache[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

95Choose the cache duration when a conversation starts. In Anthropic’s test, the first turn added at 1 hour to a conversation cached at 5 minutes wrote the whole prompt again, once: $0.80 USD at 100,000 tokens on Claude Opus 5.5, compared with $0.02 USD of input for a turn that read the cache[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

9696

97Anthropic measured a small agent that triages bug reports ([Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens)) with pauses inserted before some turns to simulate a person’s delay[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). Accuracy stayed within run-to-run noise at every setting measured. Anthropic also measured extra requests that keep the 5-minute cache warm (keep-alive requests).

97Anthropic measured a small agent that triages bug reports ([Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens)) with pauses inserted before some turns to simulate a person’s delay[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). Accuracy stayed within run-to-run noise at every setting measured. Anthropic also measured extra requests that keep the 5-minute cache warm (keep-alive requests).

9898

9999To decide, count the gaps between consecutive requests in a conversation, and find your row in the following table:

100100

from line 125

125125

126126* **Claude Fable 5.1**, whose [cache reads](https://platform.claude.com/docs/en/about-claude/pricing#prompt-caching) cost 0.025x the input price.

127127

128 * **Pauses run for minutes.** On Claude Fable 5.1, keep-alive requests[19](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) cost 13% to 20% less than the 1-hour cache, with 10% of turns paused and with a 6-minute pause before every turn.

128 * **Pauses run for minutes.** On Claude Fable 5.1, keep-alive requests[19](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) cost 13% to 20% less than the 1-hour cache, with 10% of turns paused and with a 6-minute pause before every turn.

129129 * **Pauses run toward an hour.** On Claude Fable 5.1, only with a 45-minute pause before every turn did the 1-hour cache win, by about $0.12 USD on a session of 5 bug reports.

130130

131131The following chart plots cost per session against the share of paused turns on Claude Sonnet 5.5:

from line 152

1521522. Re-send the request’s headers with its body. Without the original request’s `anthropic-beta` header, the beta-gated fields in the replayed body are rejected.

1531533. Send it within 4 minutes of the previous request’s start, and every 4 minutes after that, until the next turn arrives or the session ends. Count from the request’s start, because the [cache’s 5-minute lifetime](https://platform.claude.com/docs/en/build-with-claude/prompt-caching#how-prompt-caching-works) runs from there.

154154

155That is the [pre-warming request](https://platform.claude.com/docs/en/build-with-claude/prompt-caching#pre-warming-the-cache): it refreshes the cache’s lifetime, returns no output, and costs little more than the cache read. Once the cache has expired, the next request, keep-alive or not, re-writes the prefix at the cache write price[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

155That is the [pre-warming request](https://platform.claude.com/docs/en/build-with-claude/prompt-caching#pre-warming-the-cache): it refreshes the cache’s lifetime, returns no output, and costs little more than the cache read. Once the cache has expired, the next request, keep-alive or not, re-writes the prefix at the cache write price[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

156156

157157A `max_tokens: 0` request is rejected in the following cases:

158158

from line 190

190190

191191Keep per-request text in the newest user turn, after the last cache breakpoint, and make every cache-invalidating change at a natural break, such as the first request after a compaction. The cache is a byte-exact prefix match over the request in order (tools, then system prompt, then messages), so a change anywhere invalidates everything after it.

192192

193When something must change mid-conversation, send it in the form that keeps the cache, where the model supports it. In Anthropic’s test on Claude Fable 5.1, Claude Opus 5.5, and Claude Sonnet 5.5, on a cached conversation of about 100,000 tokens, editing the system prompt cost about 25 to 50 times what the same text cost as a mid-conversation system message. Changing effort at the top level cost 20 to 40 times what it cost per message[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs):

193When something must change mid-conversation, send it in the form that keeps the cache, where the model supports it. In Anthropic’s test on Claude Fable 5.1, Claude Opus 5.5, and Claude Sonnet 5.5, on a cached conversation of about 100,000 tokens, editing the system prompt cost about 25 to 50 times what the same text cost as a mid-conversation system message. Changing effort at the top level cost 20 to 40 times what it cost per message[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs):

194194

195195<Frame>

196196 ![Bar chart: system prompt edit up to $1.26 USD, top-level effort change up to $1.02 USD, either in a message up to $0.026 USD](https://platform.claude.com/docs/images/cost-intel/cache-break-cost.svg)

197197</Frame>

198198

199At list prices, the stakes are highest on Claude Fable 5.1 and Claude Mythos 5.1, where a break rewrites the prefix at 1.25x the input price instead of reading it at 0.025x. On Claude Haiku 5.5, a system prompt edit cost 12.5x as much as an unchanged request, because the edit rewrote the whole conversation. On models that run the [prefix check](https://platform.claude.com/docs/en/build-with-claude/preserved-thinking#prefix-check), a request that makes this edit and sends earlier thinking blocks back can get a 400 error instead[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). The top-level effort change, in a conversation that held the model’s first reply with its thinking, rewrote all but the system prompt. These figures are from the Messages API.

199At list prices, the stakes are highest on Claude Fable 5.1 and Claude Mythos 5.1, where a break rewrites the prefix at 1.25x the input price instead of reading it at 0.025x. On Claude Haiku 5.5, a system prompt edit cost 12.5x as much as an unchanged request, because the edit rewrote the whole conversation. On models that run the [prefix check](https://platform.claude.com/docs/en/build-with-claude/preserved-thinking#prefix-check), a request that makes this edit and sends earlier thinking blocks back can get a 400 error instead[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). The top-level effort change, in a conversation that held the model’s first reply with its thinking, rewrote all but the system prompt. These figures are from the Messages API.

200200

201201**What to do.**

202202

203* Put nothing that changes per request, such as a timestamp or a queue position, ahead of the stable prefix: it turns every request into a full cache write. In Anthropic’s test, a 25-token status line at the front of the system prompt made each request write the whole conversation; moved to the end of the newest user turn, it left the cache read intact[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

203* Put nothing that changes per request, such as a timestamp or a queue position, ahead of the stable prefix: it turns every request into a full cache write. In Anthropic’s test, a 25-token status line at the front of the system prompt made each request write the whole conversation; moved to the end of the newest user turn, it left the cache read intact[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

204204* The [prompt caching](https://platform.claude.com/docs/en/build-with-claude/prompt-caching#what-invalidates-the-cache) page lists the changes that invalidate the cache, such as a change to [`effort`](https://platform.claude.com/docs/en/build-with-claude/effort), the thinking configuration, the [`speed`](https://platform.claude.com/docs/en/build-with-claude/fast-mode) setting, the system prompt, or a tool definition. Changing the output format invalidates the cache for that conversation (see [structured outputs](https://platform.claude.com/docs/en/build-with-claude/structured-outputs#prompt-modification-and-token-costs)).

205205* Check which models support each form: [mid-conversation system messages](https://platform.claude.com/docs/en/build-with-claude/mid-conversation-system-messages), and [per-message effort](https://platform.claude.com/docs/en/build-with-claude/effort#changing-effort-mid-conversation), which is in beta.

206206* After a cache-invalidating change, confirm cache reads have not dropped; if they have, [cache diagnostics](https://platform.claude.com/docs/en/build-with-claude/cache-diagnostics) shows where the prefix diverged.

207207

208On an earlier model, Anthropic measured when to make such a change, on the triage agent’s long sessions[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). An effort change and an added tool cost more mid-session, or on the request that triggered the compaction, than on the first request after compaction. Accuracy was within run-to-run noise. The sessions with no changes cost more than those changed after compaction for another reason: they ran at the default effort throughout, and two of them compacted a second time, while the others started at low effort:

208On an earlier model, Anthropic measured when to make such a change, on the triage agent’s long sessions[18](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). An effort change and an added tool cost more mid-session, or on the request that triggered the compaction, than on the first request after compaction. Accuracy was within run-to-run noise. The sessions with no changes cost more than those changed after compaction for another reason: they ran at the default effort throughout, and two of them compacted a second time, while the others started at low effort:

209209

210210<Frame>

211211 ![Bar chart, cost per triage session: $0.81 USD no changes, $0.95 USD mid-session, $0.92 USD at compaction, $0.75 USD after](https://platform.claude.com/docs/images/cost-intel/compaction-timing.svg)

from line 227

227227

228228The levers interact with the cache and each other, so judge them by net effect, and use [cache diagnostics](https://platform.claude.com/docs/en/build-with-claude/cache-diagnostics) to confirm your cached prefix survives each change.

229229

230In measurements on an earlier model[28](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Anthropic ran them on an issue-triage agent working through 20 real bug reports with screenshots from a public repository, and on a longer variant of the same job with 2.6 times the tokens. With caching on, input trimming (image resizing and tool search) cut the cost of both runs further.

230In measurements on an earlier model[28](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Anthropic ran them on an issue-triage agent working through 20 real bug reports with screenshots from a public repository, and on a longer variant of the same job with 2.6 times the tokens. With caching on, input trimming (image resizing and tool search) cut the cost of both runs further.

231231

232232#### Defer unused tool definitions

233233

234234Every tool definition attached to a request is input on every turn, and a few MCP servers add up to hundreds of them.

235235

236In a measurement on an earlier model[27](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Anthropic ran the triage agent with its own two tools plus a catalog of real tool definitions from public MCP servers, for a total of up to 502 tools, loading all of them or marking the extras `defer_loading` behind [tool search](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool):

236In a measurement on an earlier model[27](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Anthropic ran the triage agent with its own two tools plus a catalog of real tool definitions from public MCP servers, for a total of up to 502 tools, loading all of them or marking the extras `defer_loading` behind [tool search](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool):

237237

238238<Frame>

239239 ![Line chart: with all tools loaded, run cost rises from $0.55 USD to $1.02 USD at 502 tools; tool search keeps it at $0.56 USD](https://platform.claude.com/docs/images/cost-intel/tool-search.svg)

from line 247

247247

248248#### Keep data files out of the prompt

249249

250When the model has to compute over a table, upload it with the [Files API](https://platform.claude.com/docs/en/build-with-claude/files) and let the model query it with [code execution](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool) instead of pasting it in. Anthropic asked 25 aggregate questions[15](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) (sums, filtered counts, group-bys, and a date filter) over a 1,862-row public CSV, with the answers computed by pandas:

250When the model has to compute over a table, upload it with the [Files API](https://platform.claude.com/docs/en/build-with-claude/files) and let the model query it with [code execution](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool) instead of pasting it in. Anthropic asked 25 aggregate questions[15](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) (sums, filtered counts, group-bys, and a date filter) over a 1,862-row public CSV, with the answers computed by pandas:

251251

252252<Frame>

253253 ![Scatter chart: file uploaded with code execution, 25 of 25 correct at $0.40 USD; pasted into the prompt, 7 of 25 at $5.01 USD](https://platform.claude.com/docs/images/cost-intel/data-files.svg)

from line 263

263263 ![Bar chart by run length: context editing adds 73% on the short run; compaction saves 32% and pruning 39% on the long](https://platform.claude.com/docs/images/cost-intel/hygiene.svg)

264264</Frame>

265265

266Anthropic measured these levers on an earlier model[28](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs):

266Anthropic measured these levers on an earlier model[28](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs):

267267

268268* On the 20-issue run the levers saved little or nothing, and context editing made the run cost more.

269269* On the long run the prune and compaction both cut the cost, while context editing changed nothing.

from line 359

359359

360360The command then proposes its edits as a diff (one hunk shown) and lists what it deliberately left alone: the refund window, the tone requirement, and the quality bar. You review a patch, not a rewrite.

361361

362Anthropic measured the effect on a support-desk evaluation[14](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), across two migrations between earlier models:

362Anthropic measured the effect on a support-desk evaluation[14](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), across two migrations between earlier models:

363363

364364<Frame>

365365 ![Scatter chart, support-desk evaluation: the old prompt costs more on the new model; audited, it is cheaper and as accurate](https://platform.claude.com/docs/images/cost-intel/prompt-audit.svg)

from line 388

388388

389389Compare models on cost per completed task, not per token, because you pay for completed tasks. For most agent workloads, start with Claude Opus 5.5 at its default effort (`medium`).

390390

391A more capable model can finish a task with less work, which can offset much of the per-token premium. Anthropic measured this on the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset, whose scores are not comparable to the public leaderboard, priced as a customer is billed:

391A more capable model can finish a task with less work, which can offset much of the per-token premium. Anthropic measured this on the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset, whose scores are not comparable to the public leaderboard, priced as a customer is billed:

392392

393393<Frame>

394394 ![Scatter chart, SWE-bench Pro: Claude Opus 5.5 at low effort matches Sonnet 5.5's default for about 7% more per solved task](https://platform.claude.com/docs/images/cost-intel/cost-per-task-sonnet-5-5.svg)

from line 396

396396

397397The following table lists the results:

398398

400| ----------------------------------------------------------------------------------------------------------------------------------------------- | ----------------- | ------------------ | --------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------- |

407| GPQA Diamond[9](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Haiku 5.5 | `medium` (default) | 85% accuracy, compared with 91% for each of Claude Opus 5.5 and Claude Sonnet 5.5 | About a twentieth of Claude Opus 5.5’s cost per correct answer and about a tenth of Claude Sonnet 5.5’s |

400| ---------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------- | ------------------ | --------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------- |

407| GPQA Diamond[9](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Haiku 5.5 | `medium` (default) | 85% accuracy, compared with 91% for each of Claude Opus 5.5 and Claude Sonnet 5.5 | About a twentieth of Claude Opus 5.5’s cost per correct answer and about a tenth of Claude Sonnet 5.5’s |

409409

410410**What to do.**

411411

from line 412

412412* Start with Claude Opus 5.5 at its default effort: on the SWE-bench Pro subset it matched Claude Fable 5.1 at its default, inside run-to-run noise, for about a fifth of the cost per solved task. Opus 5.5 and Fable 5.1 both largely saturate the SWE-bench Pro subset. Use Claude Fable 5.1 for demanding reasoning and long-horizon agentic work, or when your evals on Claude Opus 5.5 at higher effort still fall short.

413413* At the other end, Claude Haiku 5.5 suits high-volume and latency-sensitive work with checkable outputs.

414414* Price every candidate in cost per completed task on your own traffic, including Claude Opus 5.5 at its default effort and the frontier model at reduced effort. Each of Claude Opus 5.5’s token prices, cache reads included, is twice Claude Sonnet 5.5’s, but on the SWE-bench Pro subset Opus 5.5 at `low` solved about the same share of tasks as Sonnet 5.5 at its default, inside run-to-run noise, for about 7% more per solved task.

415* Measure what a more expensive model buys on your own workload. On the 45 DeepResearch Bench II[7](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) tasks scored for every model, at default effort, Claude Opus 5.5 scored 2.6 points above Claude Sonnet 5.5 for 1.5 times the cost per task, and Claude Fable 5.1 scored 6.0 points above Claude Sonnet 5.5 for 5.8 times. On all 50 tasks, with four biology tasks that only Claude Sonnet 5.5 answered counted as failures, neither is measurably ahead of Claude Sonnet 5.5.

415* Measure what a more expensive model buys on your own workload. On the 45 DeepResearch Bench II[7](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) tasks scored for every model, at default effort, Claude Opus 5.5 scored 2.6 points above Claude Sonnet 5.5 for 1.5 times the cost per task, and Claude Fable 5.1 scored 6.0 points above Claude Sonnet 5.5 for 5.8 times. On all 50 tasks, with four biology tasks that only Claude Sonnet 5.5 answered counted as failures, neither is measurably ahead of Claude Sonnet 5.5.

416416* Price the tail of your workload, not the median: compare models on the hardest tenth of your tasks, not the typical one.

417417

418The bill is decided by the tasks the cheaper model fails, because a failed task still bills its tokens, then the retry, then whatever the failure costs downstream. The tail is also where the money goes even when nothing fails. On a 20-problem WideSearch[1](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) run, two problems carried 43% of the spend:

418The bill is decided by the tasks the cheaper model fails, because a failed task still bills its tokens, then the retry, then whatever the failure costs downstream. The tail is also where the money goes even when nothing fails. On a 20-problem WideSearch[1](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) run, two problems carried 43% of the spend:

419419

420420<Frame>

421421 ![Bar chart of 20 WideSearch problems ranked by cost: the top two carry 43% of spend and the cheapest half 10%](https://platform.claude.com/docs/images/cost-intel/tail.svg)

from line 425

425425

426426### Upgrade the model

427427

428If you are a model or two behind, the cheapest lever is the model string. In Anthropic’s measurements, each newer model solved at least as many tasks as the one before it, usually for less per solved task. The chart plots recent Claude models on two benchmarks in which an agent works alone on one written task and tests check the result: the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset, where every model ran through the same test setup, and Terminal-Bench 3[20](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). The table lists what each upgrade changed:

428If you are a model or two behind, the cheapest lever is the model string. In Anthropic’s measurements, each newer model solved at least as many tasks as the one before it, usually for less per solved task. The chart plots recent Claude models on two benchmarks in which an agent works alone on one written task and tests check the result: the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset, where every model ran through the same test setup, and Terminal-Bench 3[20](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). The table lists what each upgrade changed:

429429

430430<Frame>

431431 ![Charts: on SWE-bench Pro, Opus 5.5 and Sonnet 5.5 cut cost per solved task most; on Terminal-Bench 3, Opus drops to $28 USD](https://platform.claude.com/docs/images/cost-intel/upgrade-ladder-sonnet-5-5.svg)

432432</Frame>

433433

435| ------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------ | --------------------------------------- | ----------------------------------------------------- | ------------------------------------------------------------------------------------------------ |

442| Claude Managed Agents task set[25](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Haiku 4.5 to Claude Haiku 5.5 | As served; `medium` on Claude Haiku 5.5 | From 62% to 93% of attempts passed, one run per model | From $0.041 USD to about $0.010 USD in token cost per attempt; about 84% less per passed attempt |

435| ------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------ | --------------------------------------- | ----------------------------------------------------- | ------------------------------------------------------------------------------------------------ |

442| Claude Managed Agents task set[25](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Haiku 4.5 to Claude Haiku 5.5 | As served; `medium` on Claude Haiku 5.5 | From 62% to 93% of attempts passed, one run per model | From $0.041 USD to about $0.010 USD in token cost per attempt; about 84% less per passed attempt |

443443

444444**When it pays.**

445445

446446* The size of the saving depends on the work. From Claude Opus 5 to Claude Opus 5.5, it is 78% per solved task on the SWE-bench Pro subset and 40% on typical workloads ([Introducing Claude Opus 5.5](https://www.anthropic.com/news/claude-opus-5-5)). From Claude Sonnet 5 to Claude Sonnet 5.5, it is 87% on the same subset and up to 30% for most work ([Introducing Claude Sonnet 5.5](https://www.anthropic.com/news/claude-sonnet-5-5)).

447447* Where older models already solve most tasks, the cost of each attempt decides the saving: on the SWE-bench Pro subset, Claude Opus 5.5 scores about the same as Claude Opus 5 for 78% less per solved task.

448* Where older models mostly fail, the pass rate decides the saving: on Terminal-Bench 3[20](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), the upgrade from Claude Opus 4.8 to Claude Opus 5 raises tasks solved from 15% to 41% and costs 56% less per solved task.

448* Where older models mostly fail, the pass rate decides the saving: on Terminal-Bench 3[20](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), the upgrade from Claude Opus 4.8 to Claude Opus 5 raises tasks solved from 15% to 41% and costs 56% less per solved task.

449449* Compare on cost per solved task, not per token: the same text costs about 30% more tokens on Claude 4.7 and later models, so a per-token comparison makes the newer models look more expensive by construction.

450450

451451### Tune effort

from line 454

454454

455455The following table lists Anthropic’s measurements on two benchmarks with current models; scores and costs are not comparable between the two benchmarks. Costs are per task, not per solved task as in [Compare models on cost per task](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#compare-models-on-cost-per-task). Each score pools two runs on SWE-bench Pro and three on DeepResearch Bench II, unless its cell says otherwise:

456456

458| --------------------------------------------------------------------------------------------------------------------------- | ---------------- | ------------------ | ------------------------------------ | ------------- |

459| SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset | Claude Opus 5.5 | `low` | 87.4% of tasks solved | $0.11 USD |

463| DeepResearch Bench II[7](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Fable 5.1 | `low` | Rubric score 60.5% | $3.47 USD |

458| -------------------------------------------------------------------------------------------------------------------------------------- | ---------------- | ------------------ | ------------------------------------ | ------------- |

459| SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset | Claude Opus 5.5 | `low` | 87.4% of tasks solved | $0.11 USD |

463| DeepResearch Bench II[7](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Claude Fable 5.1 | `low` | Rubric score 60.5% | $3.47 USD |

466466

467The following chart plots accuracy against cost per task for Claude Opus 5.5 on the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset:

467The following chart plots accuracy against cost per task for Claude Opus 5.5 on the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset:

468468

469469<Frame>

470470 ![Line chart of accuracy against cost per task by effort: steep from low to high for Claude Opus 5.5 on SWE-bench Pro](https://platform.claude.com/docs/images/cost-intel/effort-sweep-opus-5-5-swe-bench-pro.svg)

471471</Frame>

472472

473The following chart plots rubric score against cost per task for Claude Fable 5.1 on DeepResearch Bench II[7](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs):

473The following chart plots rubric score against cost per task for Claude Fable 5.1 on DeepResearch Bench II[7](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs):

474474

475475<Frame>

476476 ![Line chart, DeepResearch Bench II, Fable 5.1: rubric score 60.5% at low effort, 60.5% at medium, 63.0% at high, as cost rises](https://platform.claude.com/docs/images/cost-intel/effort-limit-repriced.svg)

from line 479

479479**What to do.**

480480

481481* Sweep two or three effort levels on a sample of your own traffic, each in a separate session: changing top-level effort mid-session [invalidates the cache](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#cache-repeated-context) and distorts the comparison. The curve is the baseline any multi-model strategy must beat ([step 2 of measuring on your own workload](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#measure-on-your-own-workload)).

482* On long-horizon coding, effort is a real tradeoff, which [re-running failures at higher effort](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#re-run-failures-at-higher-effort) turns back into a saving. At `medium`, Claude Sonnet 5.5 scored about 7 points lower than at its default, `high`, on the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset for about 59% of the cost per task (2 runs each), and scored 62.7% compared with 77.8% on the internal agentic-coding benchmark[11](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) (1 attempt per task).

482* On long-horizon coding, effort is a real tradeoff, which [re-running failures at higher effort](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#re-run-failures-at-higher-effort) turns back into a saving. At `medium`, Claude Sonnet 5.5 scored about 7 points lower than at its default, `high`, on the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset for about 59% of the cost per task (2 runs each), and scored 62.7% compared with 77.8% on the internal agentic-coding benchmark[11](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) (1 attempt per task).

483483* At `xhigh` and `max`, Claude Sonnet 5.5 can start its own rounds of review after it finishes a task, which takes more time and tokens. To keep that work on the task, add the prompt under “Thoroughness at `xhigh` and `max` effort” in [Prompting Claude Sonnet 5.5](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-sonnet-5-5#steer-initiative-and-scope) to your system prompt.

484* On research tasks, measure the model you ship. Claude Fable 5.1 scored 2.5 points higher at `high` than at `low` on DeepResearch Bench II, for 1.6 times the cost, and no higher at `medium` than at `low`; Claude Opus 5.5 scored 13.3 points lower at `low` than at `medium` on DRACO[21](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). The two results come from different models on different benchmarks, so they show that the answer varies, not why.

485* At `low`, Claude Opus 5.5’s typical task took 84% less time than at `medium` on DRACO[21](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), 30% less on HLE[22](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), and 47% less on an internal set of physics problems[23](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), and it scored 13.3, 3.4, and 14.0 points lower (3, 3, and 4 attempts per task).

486* On an internal defect-search benchmark for work larger than one context window[8](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Claude Fable 5.1 took 15.2, 17.5, and 19.9 hours per run at `low`, `medium`, and `high` (3 runs each). To save time, also compare lower effort with the two changes in [Show the model elapsed time](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#show-the-model-elapsed-time): `low` effort saved more time, and the changes kept more of the score.

484* On research tasks, measure the model you ship. Claude Fable 5.1 scored 2.5 points higher at `high` than at `low` on DeepResearch Bench II, for 1.6 times the cost, and no higher at `medium` than at `low`; Claude Opus 5.5 scored 13.3 points lower at `low` than at `medium` on DRACO[21](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs). The two results come from different models on different benchmarks, so they show that the answer varies, not why.

485* At `low`, Claude Opus 5.5’s typical task took 84% less time than at `medium` on DRACO[21](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), 30% less on HLE[22](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), and 47% less on an internal set of physics problems[23](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), and it scored 13.3, 3.4, and 14.0 points lower (3, 3, and 4 attempts per task).

486* On an internal defect-search benchmark for work larger than one context window[8](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Claude Fable 5.1 took 15.2, 17.5, and 19.9 hours per run at `low`, `medium`, and `high` (3 runs each). To save time, also compare lower effort with the two changes in [Show the model elapsed time](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#show-the-model-elapsed-time): `low` effort saved more time, and the changes kept more of the score.

487487

488488### Re-run failures at higher effort

489489

490490When a task’s outcome is checkable, the cheapest policy on the effort curve is not a fixed setting: run every task at a low setting and re-run only the failures at a higher one.

491491

492Anthropic computed this policy task by task from the Claude Opus 5.5 effort runs on the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset in [Tune effort](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#tune-effort):

492Anthropic computed this policy task by task from the Claude Opus 5.5 effort runs on the SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) subset in [Tune effort](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#tune-effort):

493493

494494<Frame>

495495 ![Chart, SWE-bench Pro, Opus 5.5: low or medium effort with failures re-run at high matches any fixed effort for less than high](https://platform.claude.com/docs/images/cost-intel/escalation-opus-5-5.svg)

from line 518

518518* Set `max_tokens` to 64,000 for agentic work, or to 128,000, the maximum, when a single cut-off attempt is costly. [Stream responses](https://platform.claude.com/docs/en/build-with-claude/streaming) that large, treat [`stop_reason: max_tokens`](https://platform.claude.com/docs/en/build-with-claude/handling-stop-reasons#max-tokens) as a failure, and save money with effort, which the model can see.

519519* On Claude Haiku 5.5, raise a small `max_tokens` or lower effort: thinking is on by default and counts toward the cap, so a response can end before any text (see the [Claude Haiku 5.5 migration guide](https://platform.claude.com/docs/en/models/haiku-5-5/migration-guide#configure-thinking)).

520520

521The following table shows what each model solved and what a solved task cost at each cap on an internal repository-task benchmark[12](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), one attempt per task, each model at its default effort. The rows at 16,384 average two runs, and the other rows are one run each:

521The following table shows what each model solved and what a solved task cost at each cap on an internal repository-task benchmark[12](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), one attempt per task, each model at its default effort. The rows at 16,384 average two runs, and the other rows are one run each:

522522

524524| ---------------- | ------------ | ------------------------- | ------------ | ------------------------ |

from line 533

533533

534534Compare the caps within each model. The two models were run a month apart, and their costs are priced on different bases.

535535

536On the internal repository tasks[12](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), almost every turn finishes far below a 16,384-token cap. The rare long turn is what a higher cap buys:

536On the internal repository tasks[12](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), almost every turn finishes far below a 16,384-token cap. The rare long turn is what a higher cap buys:

537537

538538<Frame>

539539 ![Dot plot of per-turn output for Opus 5.5 and Fable 5.1: medians a few hundred tokens, longest 61k and 128k, against the caps](https://platform.claude.com/docs/images/cost-intel/max-tokens-ladder-opus-5-5.svg)

from line 541

541541

542542Ask for shorter answers. Output tokens cost five times input tokens on Claude Haiku 5.5, Claude Sonnet 5.5, Claude Opus 5.5, and Claude Fable 5.1, and in an agent loop every token the model writes comes back as input on every later turn, so you pay for a long answer again and again.

543543

544In a measurement on an earlier model[26](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Anthropic ran a small agent that triages bug reports ([Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens)) under three final-answer instructions, three runs each, with the same model and tools:

544In a measurement on an earlier model[26](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Anthropic ran a small agent that triages bug reports ([Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens)) under three final-answer instructions, three runs each, with the same model and tools:

545545

546546* The original asked for two lines.

547547* The shorter variant asked for one.

from line 557

557557

558558### Show the model elapsed time

559559

560To make an agent finish sooner and cost less, tell it that time matters, and show it the elapsed time on every request. In Anthropic’s measurement, the two changes cut the time of a single Claude Opus 5.5 agent at `medium` effort by 47% and its cost by 60% on the typical (median) DRACO[21](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) research task, and its average score was 4.1 points lower.

560To make an agent finish sooner and cost less, tell it that time matters, and show it the elapsed time on every request. In Anthropic’s measurement, the two changes cut the time of a single Claude Opus 5.5 agent at `medium` effort by 47% and its cost by 60% on the typical (median) DRACO[21](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) research task, and its average score was 4.1 points lower.

561561

562On HLE[22](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) and an internal physics set[23](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), the typical task took 14% and 7% less time, and the physics score was 3.0 points lower. Lowering effort to `low` saved more time and cost than the changes did, and lost more score (reference 21). Check the score on your own tasks before you adopt the changes.

562On HLE[22](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) and an internal physics set[23](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), the typical task took 14% and 7% less time, and the physics score was 3.0 points lower. Lowering effort to `low` saved more time and cost than the changes did, and lost more score (reference 21). Check the score on your own tasks before you adopt the changes.

563563

564564For working code, see the Claude Cookbook recipe [Multi-agent teams under latency pressure and budgets](https://github.com/anthropics/claude-cookbooks/blob/main/patterns/agents/latency_multi_agent.ipynb), which applies both changes to a team of agents. To say that time matters in the system prompt instead, see [Time signals for multiagent harnesses](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-opus-5-5#time-signals-for-multi-agent-harnesses).

565565

from line 596

596596

597597The following table compares each pairing with its executor alone at the same effort:

598598

600| ---------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------- | -------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------- |

601| Internal agentic-coding benchmark[11](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | 90.1%, 1.7 points over the executor alone, at the edge of run-to-run noise | $2.92 USD per attempt, about 2.1 times the executor alone | About 1.4 times per attempt; 4% of attempts received no advice |

602| GPQA Diamond[9](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Within run-to-run noise of the executor alone | 12% more per question for Claude Haiku 5.5, 25% more for Claude Sonnet 5.5 | On none of the 198 questions, in two runs each; no system prompt asked for consultations |

603| Chartography[13](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | 61.7, 7 points below the executor alone and beyond run-to-run noise | About the same as the executor alone | On 1 of 300 tasks |

600| --------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------- | -------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------- |

601| Internal agentic-coding benchmark[11](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | 90.1%, 1.7 points over the executor alone, at the edge of run-to-run noise | $2.92 USD per attempt, about 2.1 times the executor alone | About 1.4 times per attempt; 4% of attempts received no advice |

602| GPQA Diamond[9](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Within run-to-run noise of the executor alone | 12% more per question for Claude Haiku 5.5, 25% more for Claude Sonnet 5.5 | On none of the 198 questions, in two runs each; no system prompt asked for consultations |

603| Chartography[13](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | 61.7, 7 points below the executor alone and beyond run-to-run noise | About the same as the executor alone | On 1 of 300 tasks |

604604

605605* **Internal agentic-coding benchmark.** Claude Opus 5.5 at `high`, with a Claude Fable 5.1 advisor.

606606* **GPQA Diamond.** Claude Haiku 5.5 or Claude Sonnet 5.5 at each model’s default (`medium` and `high`), with a Claude Opus 5.5 advisor.

from line 617

617617**What to do.**

618618

619619* Measure the consult rate on your own tasks. The advisor tool’s definition, about 1,000 prompt tokens per request, adds cost even when the executor never consults. With [prompt caching](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#cache-repeated-context), a multi-turn agent loop pays for the definition at the cache-read price after the first request, so this overhead matters most for short, single-request tasks.

620* If the executor rarely consults, see [Prompting for coding and agent tasks](https://platform.claude.com/docs/en/agents-and-tools/tool-use/advisor-tool#prompting-for-coding-and-agent-tasks) for system prompts that ask for consultations, and measure again after you add one. On 20 SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) problems, the suggested prompt made no clear difference for Claude Sonnet 5.5.

621* Restore the executor’s effort if the consult rate collapses. In measurements on earlier models, one pairing that consulted on most tasks at the default effort consulted on few when effort was lowered, and then scored below the executor alone at its default effort (SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs)).

620* If the executor rarely consults, see [Prompting for coding and agent tasks](https://platform.claude.com/docs/en/agents-and-tools/tool-use/advisor-tool#prompting-for-coding-and-agent-tasks) for system prompts that ask for consultations, and measure again after you add one. On 20 SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) problems, the suggested prompt made no clear difference for Claude Sonnet 5.5.

621* Restore the executor’s effort if the consult rate collapses. In measurements on earlier models, one pairing that consulted on most tasks at the default effort consulted on few when effort was lowered, and then scored below the executor alone at its default effort (SWE-bench Pro[3](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs)).

622622* Recheck at every model release, because releases move both the capability gap and the price ratio.

623623

624624### Orchestrator strategy: delegate bulk work

from line 625

625625

626626Delegation saved money in only two situations, in measurements with workers on Claude Sonnet 5, an earlier model:

627627

628* Routine work with a long cost tail (an easy slice of BrowseComp[4](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs))

629* Work larger than one context window (an internal defect-search benchmark[8](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs))

628* Routine work with a long cost tail (an easy slice of BrowseComp[4](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs))

629* Work larger than one context window (an internal defect-search benchmark[8](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs))

630630

631When the work is one dependent chain, or fits in a single context without a long cost tail, the orchestrator’s model alone at lower effort came out ahead in every case measured on earlier models (the harder BrowseComp set and DeepWideSearch[6](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs)); independent external work reports the same pattern[5](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

631When the work is one dependent chain, or fits in a single context without a long cost tail, the orchestrator’s model alone at lower effort came out ahead in every case measured on earlier models (the harder BrowseComp set and DeepWideSearch[6](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs)); independent external work reports the same pattern[5](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

632632

633633To build an orchestrator, use [multiagent orchestration](https://platform.claude.com/docs/en/managed-agents/multiagent-orchestration) in Claude Managed Agents: configure an orchestrator agent and worker agents that it can call, each with its own model.

634634

from line 645

645645**What to do.**

646646

647647* **Routine work with a long cost tail:** measure this case first, for accuracy as well as cost. The measurement behind it covers only problems that the delegated configuration answered correctly in every run (reference 4).

648* **Work larger than one context window:** delegate partitions to workers. Lowering effort cannot help, because the bill is the reading itself: on the defect-search benchmark[8](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Claude Fable 5.1 alone cost $468 USD to $552 USD per run across its three effort settings and took 15 to 20 hours. With earlier-model workers, delegation cost less and finished sooner, for a lower score.

649* **Claude Haiku 5.5 subagents on coding tasks:** compare them with lower effort first. On an internal coding benchmark on Claude Managed Agents, they were an alternative to lowering effort, not better. Adding them to Claude Opus 5.5 at `medium`, with a short delegation prompt, cut cost per attempt by 32% and the pass rate by about 4 percentage points (95% interval 1.3 to 7.1), about what a mix of `low` and `medium` without subagents would give at that cost[29](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

648* **Work larger than one context window:** delegate partitions to workers. Lowering effort cannot help, because the bill is the reading itself: on the defect-search benchmark[8](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), Claude Fable 5.1 alone cost $468 USD to $552 USD per run across its three effort settings and took 15 to 20 hours. With earlier-model workers, delegation cost less and finished sooner, for a lower score.

649* **Claude Haiku 5.5 subagents on coding tasks:** compare them with lower effort first. On an internal coding benchmark on Claude Managed Agents, they were an alternative to lowering effort, not better. Adding them to Claude Opus 5.5 at `medium`, with a short delegation prompt, cut cost per attempt by 32% and the pass rate by about 4 percentage points (95% interval 1.3 to 7.1), about what a mix of `low` and `medium` without subagents would give at that cost[29](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs).

650650

651651## Measure on your own workload

652652

from line 853

853853

854854The following table lists the levers in the order to try them. The figures come from Anthropic’s tests, launch posts, and published prices and speeds:

855855

857| -------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

858| Prompt caching | Cost cut by a factor of 4.8 to 7.2 on a research agent, and by 83% to 86% on a small triage agent | None | Faster | [Cache repeated context](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#cache-repeated-context) |

859| 1-hour cache duration | With 1 turn in 10 after a pause, the 5-minute default cost 54% more; with 1 in 20, about the same | None | On earlier models, stayed warm after a pause[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | [Pick the cache duration](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#pick-the-cache-duration) |

860| Input trimming | A further 21% to 26% off a small triage agent’s bill | With tool search, accuracy in the same range as without[27](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Neutral | [Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens) |

861| Tool search | 45% with 500 tool definitions attached; 20% from deferring a GitHub MCP server’s toolset | None for tools the task never needs | Neutral | [Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens) |

862| Data files through code execution | 92% on a data task of 25 questions, with prompt caching off | A gain: 25 of 25 correct instead of 7 | Faster | [Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens) |

864| Prompt audit against the current model | 14% on both model upgrades tested | None; a gain on one | Faster (fewer tool rounds) | [Audit prompts against the current model](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#audit-prompts-against-the-current-model) |

865| Upgrade the model | Usually less: up to 40% less on typical workloads, and up to 87% less per solved task on a coding benchmark | None; a gain on most | Neutral | [Upgrade the model](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#upgrade-the-model) |

866| Lower effort | On long coding tasks, `medium` cost about 30% to 52% less than `high`, and `low` about two thirds less | Up to 15 points lower, depending on the task; 2.5 on one research agent | Faster | [Tune effort](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#tune-effort) |

867| Re-run failures | About 40% less than running everything at `high`, at the same pass rate or slightly better | None | A second attempt on the tasks that fail | [Re-run failures at higher effort](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#re-run-failures-at-higher-effort) |

869| Raising `max_tokens` | None: cost per solved task stayed about the same | A gain: up to 22 points more tasks solved on one coding benchmark; none on another | Neutral | [Set budgets and output caps](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#set-budgets-and-output-caps) |

870| Advisor | None: about 2.1 times the cost of the model alone, on a coding benchmark | On coding, a small gain that may be chance; on chart reading, a loss | About 1.4 extra calls per attempt, on coding | [Advisor strategy](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#advisor-strategy-escalate-hard-decisions) |

871| Orchestrator | About half the cost of the orchestrator’s model alone, on work larger than one context window; on routine work, measure first | 10 to 12 points lower on that work | Much faster on that work | [Orchestrator strategy](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#orchestrator-strategy-delegate-bulk-work) |

872| Fast mode | None: twice the standard token prices | None | Up to 2.5x higher output tokens per second | [Show the model elapsed time](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#show-the-model-elapsed-time) |

857| -------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

859| 1-hour cache duration | With 1 turn in 10 after a pause, the 5-minute default cost 54% more; with 1 in 20, about the same | None | On earlier models, stayed warm after a pause[16](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | [Pick the cache duration](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#pick-the-cache-duration) |

860| Input trimming | A further 21% to 26% off a small triage agent’s bill | With tool search, accuracy in the same range as without[27](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs) | Neutral | [Trim input and context tokens](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#trim-input-and-context-tokens) |

873873

874874### Automate the search with the Claude API skill

875875

from line 889

889889/claude-api hillclimb

890890```

891891

892Anthropic measured the first lever of `cost-optimize` and two searches on tau2-bench retail[24](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), all on earlier models. The cache breakpoints that `cost-optimize` produced cut cost per task, and the pass rate fell by less than the spread of identical runs.

892Anthropic measured the first lever of `cost-optimize` and two searches on tau2-bench retail[24](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#refs), all on earlier models. The cache breakpoints that `cost-optimize` produced cut cost per task, and the pass rate fell by less than the spread of identical runs.

893893

894894A search found a cheaper configuration by lowering effort and moving down a model tier. Its pass rate stayed inside the band set before the search by a fraction of a task, and ended below the stricter floor the search used for its own verdicts. A second search, started from the lower-cost model at its default effort, kept nothing. Reference 24 has the figures and what each search cost to run.

895895

This page’s history 33 edits

4 Nov 2026 +5 −3
30 Oct 2026 +101 −101 · this one
29 Oct 2026 +32 −35
29 Oct 2026 +4 −4
28 Oct 2026 +14 −4
27 Oct 2026 +186 −186

The page’s own history →

Where it came from

The captureapi
Upstream pageplatform.claude.com

Around it

Every changethe ledger
That day30 Oct 2026