Create a Message changedapi/beta/messages/create
Nearest release: v2.1.306, 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
Lines+55added
Lines−1removed
From line
129
where the diff opens
First seen
14 Aug 2026
this site's first read of the page
Recorded edits29to this page, all time
The whole hunk
from line 129, old and new numbered
/
This page is larger than the 256 KiB this site keeps, so one side of the diff below stops where the stored text does.
from line 129
129129
130130 - `"telemetry-destinations-2026-08-11"`
131131
132 - `"webhooks-2026-10-07"`
133
132134- `"anthropic-user-profile-id": optional string`
133135
134136 The user profile ID to attribute this request to. Use when acting on behalf of a party other than your organization.
from line 3776
37743776
37753777 - `max_tokens: optional number or null`
37763778
3779 The `max_tokens` for the fallback attempt that uses this `fallbacks` entry's `model`, in place of the request's `max_tokens`. If you leave it out or set it to `null`, the attempt uses the request's `max_tokens`.
3780
3781 See the [refusals and fallback guide](https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback?sdk_languages=http#naming-your-own-fallback-models) to learn more about fallback entries.
3782
37773783 - `output_config: optional BetaOutputConfig or null`
37783784
3785 Output parameters for the fallback attempt that uses this `fallbacks` entry's `model`, combined with the request's `output_config` one field at a time.
3786
3787 - `format`: If you leave it out, the attempt uses the request's `format`. If you set it to `null`, the attempt runs without a format.
3788 - `effort`: If you leave it out, the attempt uses the request's `effort` when `model` supports that level with the attempt's `thinking`. If the request doesn't set `effort`, if `model` doesn't support the level, or if you set `effort` to `null`, the attempt runs at the default effort of `model`.
3789 - `task_budget`: If you leave it out, the attempt uses the request's `task_budget` when `model` accepts a task budget of that size. If the request doesn't set `task_budget`, if `model` doesn't accept it, or if you set `task_budget` to `null`, the attempt runs without a task budget.
3790
3791 Set `output_config` to `null` to run the attempt without a format or a task budget, at the default effort of `model`.
3792
3793 These rules apply with the `server-side-fallback-2026-07-01` beta header.
3794
37793795 - `effort: optional "low" or "medium" or "high" or 2 more or null`
37803796
37813797 How much effort the model should put into its response. Higher effort levels may result in more thorough analysis but take longer.
from line 3840
38243840
38253841 - `speed: optional "standard" or "fast" or null`
38263842
3827 Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.
3843 The inference speed mode for the fallback attempt that uses this `fallbacks` entry's `model`. If you leave it out, the attempt runs at `fast` speed only when the request sets `speed` to `fast` and your organization can use fast mode on `model`.
38283844
3845 Otherwise, or when you set this field to `null`, the attempt runs at `standard` speed.
3846
3847 These rules apply with the `server-side-fallback-2026-07-01` beta header.
3848
38293849 - `"standard"`
38303850
38313851 - `"fast"`
from line 3852
38323852
38333853 - `thinking: optional BetaThinkingConfigEnabled or BetaThinkingConfigDisabled or BetaThinkingConfigBetweenTools or BetaThinkingConfigAdaptive or null`
38343854
3855 The thinking configuration for the fallback attempt that uses this `fallbacks` entry's `model`. If you leave it out, the attempt uses the request's `thinking` when `model` supports that thinking `type`.
3856
3857 If you leave it out and the request's thinking `type` is `between_tools` but `model` doesn't support `between_tools`, the attempt uses `disabled` if `model` supports it. Otherwise, if `model` doesn't support the request's thinking `type`, if the request doesn't set `thinking`, or if you set this field to `null`, the attempt uses the thinking `type` that `model` uses when a request leaves out `thinking`. The request's `thinking.display` still applies where `model` allows it.
3858
3859 These rules apply with the `server-side-fallback-2026-07-01` beta header.
3860
38353861 - `BetaThinkingConfigEnabled object`
38363862
38373863 - `type: "enabled"`
from line 7120
70947120
70957121 - `max_tokens: optional number or null`
70967122
7097 Bounds the advisor's total output (thinking + text) per call. When the advisor hits this cap, the returned advisor_result or advisor_redacted_result block carries stop_reason='max_tokens', and a truncation note is appended to the advice text the worker model sees (inside the encrypted blob in redacted mode). When set, the server also emits a remaining-tokens budget block in the advisor's prompt so the advisor self-shapes toward the cap. When omitted, the advisor model's default output cap applies and no budget block is emitted.
7098
7099 minimum: 1024
7100
7101 - `max_uses: optional number or null`
7102
7103 Maximum number of times the tool can be used in the API request.
7104
7105 minimum: 1
7106
7107 - `strict: optional boolean`
7108
7109 When true, guarantees schema validation on tool names and inputs
7110
7111 - `BetaToolSearchToolBm25_20251119 object`
7112
7113 - `type: "tool_search_tool_bm25_20251119" or "tool_search_tool_bm25"`
7114
7115 - `"tool_search_tool_bm25_20251119"`
7116
7117 - `"tool_search_tool_bm25"`
7118
7119 - `name: "tool_search_tool_bm25"`
7120
7121 Name of the tool.
7122
7123 This is how the tool will be called by the model and in `tool_use` blocks.
7124
7125 - `allowed_callers: optional array of "direct" or "code_execution_20250825" or "code_execution_20260120" or "code_execution_20260521"`
7126
7127 - `"direct"`
7128
7129 - `"code_execution_20250825"`
7130
7131 - `"code_execution_20260120"`
7132
7133 - `"code_execution_20260521"`
7134
7135 - `cache_control: optional BetaCacheControlEphemeral or null`
7136
7137 Create a cache control breakpoint at this content block.
7138
7139 - `defer_loading: optional boolean`
7140
7141 If true, tool will not be included in initial system prompt. Only loaded when returned via tool_reference from tool search.
7142
7143 - `strict: optional boolean`
7144
7145 When true, guarantees schema validation on tool names and inputs
7146
7147 - `BetaToolSearchToolRegex20251119 object`
7148
7149 - `type: "tool_search_tool_regex_20251119" or "tool_search_tool_regex"`
7150
7151 - `"tool_search_tool_regex_20251119"`
7152
7153 - `"tool_search_tool_regex"`
7154
7155 - `name: "tool_search_tool_regex"`
7156
7157 Name of the tool.
7158
7159 This is how the tool will be called by the model and in `tool_use` blocks.
7160
7161 - `allowed_callers: optional array of "direct" or
7123 Bounds the advisor's total output (thinking + text) per call. When the advisor hits this cap, the return
No line in this hunk matches that.