@ai-sdk/anthropic 3.0.69 → 3.0.71

package/docs/05-anthropic.mdx

@@ -122,10 +122,14 @@ The following optional provider options are available for Anthropic models:
   If you are experiencing issues with the model handling requests involving
   reasoning content, you can set this to `false` to omit them from the request.
 
-- `effort` _"high" | "medium" | "low"_
+- `effort` _"low" | "medium" | "high" | "xhigh" | "max"_
 
   Optional. See [Effort section](#effort) for more details.
 
+- `taskBudget` _object_
+
+  Optional. See [Task Budgets section](#task-budgets) for more details.
+
 - `speed` _"fast" | "standard"_
 
   Optional. See [Fast Mode section](#fast-mode) for more details.
@@ -187,7 +191,7 @@ const result = streamText({
 
 ### Effort
 
-Anthropic introduced an `effort` option with `claude-opus-4-5` that affects thinking, text responses, and function calls. Effort defaults to `high` and you can set it to `medium` or `low` to save tokens and to lower time-to-last-token latency (TTLT).
+Anthropic introduced an `effort` option with `claude-opus-4-5` that affects thinking, text responses, and function calls. Effort defaults to `high` and you can set it to `medium` or `low` to save tokens and to lower time-to-last-token latency (TTLT). `claude-opus-4-7` additionally supports `xhigh` for maximum reasoning effort.
 
 ```ts highlight="8-10"
 import { anthropic, AnthropicLanguageModelOptions } from '@ai-sdk/anthropic';
@@ -228,6 +232,46 @@ const { text } = await generateText({
 
 The `speed` option accepts `'fast'` or `'standard'` (default behavior).
 
+### Task Budgets
+
+`claude-opus-4-7` supports a `taskBudget` option that informs the model of the total token budget available for an agentic turn. The model uses this information to prioritize work, plan ahead, and wind down gracefully as the budget is consumed.
+
+Task budgets are advisory — they do not enforce a hard token limit. The model will attempt to stay within budget, but actual usage may vary.
+
+```ts highlight="8-13"
+import { anthropic, AnthropicLanguageModelOptions } from '@ai-sdk/anthropic';
+import { generateText } from 'ai';
+
+const { text } = await generateText({
+  model: anthropic('claude-opus-4-7'),
+  prompt: 'Research the pros and cons of Rust vs Go for building CLI tools.',
+  providerOptions: {
+    anthropic: {
+      taskBudget: {
+        type: 'tokens',
+        total: 400000,
+      },
+    } satisfies AnthropicLanguageModelOptions,
+  },
+});
+```
+
+For long-running agents that compact and restart context, you can carry the remaining budget forward using the `remaining` field:
+
+```ts
+taskBudget: {
+  type: 'tokens',
+  total: 400000,
+  remaining: 215000, // budget left after prior compacted-away contexts
+}
+```
+
+The `taskBudget` object accepts:
+
+- `type` _"tokens"_ - Budget type. Currently only `"tokens"` is supported.
+- `total` _number_ - Total task budget for the agentic turn. Minimum 20,000.
+- `remaining` _number_ - Budget left after prior compacted-away contexts. Must be between 0 and `total`. Defaults to `total` if omitted.
+
 ### Data Residency
 
 Anthropic supports an [`inferenceGeo` option](https://platform.claude.com/docs/en/build-with-claude/data-residency) that controls where model inference runs for a request.
@@ -292,6 +336,31 @@ const { text } = await generateText({
 });
 ```
 
+##### Thinking Display (Opus 4.7+)
+
+Starting with `claude-opus-4-7`, thinking content is omitted from the response by default — thinking blocks are present in the stream but their text is empty. To receive reasoning output, set `display: 'summarized'`:
+
+```ts highlight="5"
+const { text, reasoningText } = await generateText({
+  model: anthropic('claude-opus-4-7'),
+  providerOptions: {
+    anthropic: {
+      thinking: { type: 'adaptive', display: 'summarized' },
+    } satisfies AnthropicLanguageModelOptions,
+  },
+  prompt: 'How many people will live in the world in 2040?',
+});
+
+console.log(reasoningText); // reasoning text (empty without display: 'summarized')
+console.log(text);
+```
+
+<Note>
+  If you stream reasoning to users with `claude-opus-4-7`, the default `"omitted"` display will
+  cause a long pause before output begins. Set `display: "summarized"` to restore visible
+  progress during thinking.
+</Note>
+
 #### Budget-Based Thinking
 
 For earlier models (`claude-opus-4-20250514`, `claude-sonnet-4-20250514`, `claude-sonnet-4-5-20250929`),
@@ -1376,6 +1445,7 @@ and the `mediaType` should be set to `'application/pdf'`.
 
 | Model               | Image Input         | Object Generation   | Tool Usage          | Computer Use        | Web Search          | Tool Search         | Compaction          |
 | ------------------- | ------------------- | ------------------- | ------------------- | ------------------- | ------------------- | ------------------- | ------------------- |
+| `claude-opus-4-7`   | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> |
 | `claude-opus-4-6`   | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> |
 | `claude-sonnet-4-6` | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> |                     |
 | `claude-opus-4-5`   | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> | <Check size={18} /> |                     |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-sdk/anthropic",
-  "version": "3.0.69",
+  "version": "3.0.71",
   "license": "Apache-2.0",
   "sideEffects": false,
   "main": "./dist/index.js",

package/src/anthropic-messages-language-model.ts

@@ -269,9 +269,37 @@ export class AnthropicMessagesLanguageModel implements LanguageModelV3 {
     const {
       maxOutputTokens: maxOutputTokensForModel,
       supportsStructuredOutput: modelSupportsStructuredOutput,
+      rejectsSamplingParameters,
       isKnownModel,
     } = getModelCapabilities(this.modelId);
 
+    if (rejectsSamplingParameters) {
+      if (temperature != null) {
+        warnings.push({
+          type: 'unsupported',
+          feature: 'temperature',
+          details: `temperature is not supported by ${this.modelId} and will be ignored`,
+        });
+        temperature = undefined;
+      }
+      if (topK != null) {
+        warnings.push({
+          type: 'unsupported',
+          feature: 'topK',
+          details: `topK is not supported by ${this.modelId} and will be ignored`,
+        });
+        topK = undefined;
+      }
+      if (topP != null) {
+        warnings.push({
+          type: 'unsupported',
+          feature: 'topP',
+          details: `topP is not supported by ${this.modelId} and will be ignored`,
+        });
+        topP = undefined;
+      }
+    }
+
     const isAnthropicModel = isKnownModel || this.modelId.startsWith('claude-');
 
     const supportsStructuredOutput =
@@ -345,6 +373,10 @@ export class AnthropicMessagesLanguageModel implements LanguageModelV3 {
       thinkingType === 'enabled'
         ? anthropicOptions?.thinking?.budgetTokens
         : undefined;
+    const thinkingDisplay =
+      thinkingType === 'adaptive'
+        ? anthropicOptions?.thinking?.display
+        : undefined;
 
     const maxTokens = maxOutputTokens ?? maxOutputTokensForModel;
 
@@ -364,9 +396,11 @@ export class AnthropicMessagesLanguageModel implements LanguageModelV3 {
         thinking: {
           type: thinkingType,
           ...(thinkingBudget != null && { budget_tokens: thinkingBudget }),
+          ...(thinkingDisplay != null && { display: thinkingDisplay }),
         },
       }),
       ...((anthropicOptions?.effort ||
+        anthropicOptions?.taskBudget ||
         (useStructuredOutput &&
           responseFormat?.type === 'json' &&
           responseFormat.schema != null)) && {
@@ -374,6 +408,15 @@ export class AnthropicMessagesLanguageModel implements LanguageModelV3 {
           ...(anthropicOptions?.effort && {
             effort: anthropicOptions.effort,
           }),
+          ...(anthropicOptions?.taskBudget && {
+            task_budget: {
+              type: anthropicOptions.taskBudget.type,
+              total: anthropicOptions.taskBudget.total,
+              ...(anthropicOptions.taskBudget.remaining != null && {
+                remaining: anthropicOptions.taskBudget.remaining,
+              }),
+            },
+          }),
           ...(useStructuredOutput &&
             responseFormat?.type === 'json' &&
             responseFormat.schema != null && {
@@ -612,14 +655,16 @@ export class AnthropicMessagesLanguageModel implements LanguageModelV3 {
       betas.add('effort-2025-11-24');
     }
 
+    if (anthropicOptions?.taskBudget) {
+      betas.add('task-budgets-2026-03-13');
+    }
+
     if (anthropicOptions?.speed === 'fast') {
       betas.add('fast-mode-2026-02-01');
     }
 
-    // only when streaming: enable fine-grained tool streaming
-    if (stream && (anthropicOptions?.toolStreaming ?? true)) {
-      betas.add('fine-grained-tool-streaming-2025-05-14');
-    }
+    const defaultEagerInputStreaming =
+      stream && (anthropicOptions?.toolStreaming ?? true);
 
     const {
       tools: anthropicTools,
@@ -635,6 +680,7 @@ export class AnthropicMessagesLanguageModel implements LanguageModelV3 {
             cacheControlValidator,
             supportsStructuredOutput: false,
             supportsStrictTools,
+            defaultEagerInputStreaming,
           }
         : {
             tools: tools ?? [],
@@ -643,6 +689,7 @@ export class AnthropicMessagesLanguageModel implements LanguageModelV3 {
             cacheControlValidator,
             supportsStructuredOutput,
             supportsStrictTools,
+            defaultEagerInputStreaming,
           },
     );
 
@@ -2284,15 +2331,24 @@ export class AnthropicMessagesLanguageModel implements LanguageModelV3 {
 function getModelCapabilities(modelId: string): {
   maxOutputTokens: number;
   supportsStructuredOutput: boolean;
+  rejectsSamplingParameters: boolean;
   isKnownModel: boolean;
 } {
-  if (
+  if (modelId.includes('claude-opus-4-7')) {
+    return {
+      maxOutputTokens: 128000,
+      supportsStructuredOutput: true,
+      rejectsSamplingParameters: true,
+      isKnownModel: true,
+    };
+  } else if (
     modelId.includes('claude-sonnet-4-6') ||
     modelId.includes('claude-opus-4-6')
   ) {
     return {
       maxOutputTokens: 128000,
       supportsStructuredOutput: true,
+      rejectsSamplingParameters: false,
       isKnownModel: true,
     };
   } else if (
@@ -2303,36 +2359,42 @@ function getModelCapabilities(modelId: string): {
     return {
       maxOutputTokens: 64000,
       supportsStructuredOutput: true,
+      rejectsSamplingParameters: false,
       isKnownModel: true,
     };
   } else if (modelId.includes('claude-opus-4-1')) {
     return {
       maxOutputTokens: 32000,
       supportsStructuredOutput: true,
+      rejectsSamplingParameters: false,
       isKnownModel: true,
     };
   } else if (modelId.includes('claude-sonnet-4-')) {
     return {
       maxOutputTokens: 64000,
       supportsStructuredOutput: false,
+      rejectsSamplingParameters: false,
       isKnownModel: true,
     };
   } else if (modelId.includes('claude-opus-4-')) {
     return {
       maxOutputTokens: 32000,
       supportsStructuredOutput: false,
+      rejectsSamplingParameters: false,
       isKnownModel: true,
     };
   } else if (modelId.includes('claude-3-haiku')) {
     return {
       maxOutputTokens: 4096,
       supportsStructuredOutput: false,
+      rejectsSamplingParameters: false,
       isKnownModel: true,
     };
   } else {
     return {
       maxOutputTokens: 4096,
       supportsStructuredOutput: false,
+      rejectsSamplingParameters: false,
       isKnownModel: false,
     };
   }

package/src/anthropic-messages-options.ts

@@ -17,6 +17,7 @@ export type AnthropicMessagesModelId =
   | 'claude-sonnet-4-5'
   | 'claude-sonnet-4-6'
   | 'claude-opus-4-6'
+  | 'claude-opus-4-7'
   | (string & {});
 
 /**
@@ -83,6 +84,12 @@ export const anthropicLanguageModelOptions = z.object({
       z.object({
         /** for Sonnet 4.6, Opus 4.6, and newer models */
         type: z.literal('adaptive'),
+        /**
+         * Controls whether thinking content is included in the response.
+         * - `"omitted"`: Thinking blocks are present but text is empty (default for Opus 4.7+).
+         * - `"summarized"`: Thinking content is returned. Required to see reasoning output.
+         */
+        display: z.enum(['omitted', 'summarized']).optional(),
       }),
       z.object({
         /** for models before Opus 4.6, except Sonnet 4.6 still supports it */
@@ -170,10 +177,11 @@ export const anthropicLanguageModelOptions = z.object({
     .optional(),
 
   /**
-   * Whether to enable tool streaming (and structured output streaming).
-   *
-   * When set to false, the model will return all tool calls and results
-   * at once after a delay.
+   * Whether to enable fine-grained (eager) streaming of tool call inputs
+   * and structured outputs for every function tool in the request. When
+   * true (the default), each function tool receives a default of
+   * `eager_input_streaming: true` unless it explicitly sets
+   * `providerOptions.anthropic.eagerInputStreaming`.
    *
    * @default true
    */
@@ -182,7 +190,22 @@ export const anthropicLanguageModelOptions = z.object({
   /**
    * @default 'high'
    */
-  effort: z.enum(['low', 'medium', 'high', 'max']).optional(),
+  effort: z.enum(['low', 'medium', 'high', 'xhigh', 'max']).optional(),
+
+  /**
+   * Task budget for agentic turns. Informs the model of the total token budget
+   * available for the current task, allowing it to prioritize work and wind down
+   * gracefully as the budget is consumed.
+   *
+   * Advisory only — does not enforce a hard token limit.
+   */
+  taskBudget: z
+    .object({
+      type: z.literal('tokens'),
+      total: z.number().int().min(20000),
+      remaining: z.number().int().min(0).optional(),
+    })
+    .optional(),
 
   /**
    * Enable fast mode for faster inference (2.5x faster output token speeds).

package/src/anthropic-prepare-tools.ts

@@ -27,6 +27,7 @@ export async function prepareTools({
   cacheControlValidator,
   supportsStructuredOutput,
   supportsStrictTools,
+  defaultEagerInputStreaming = false,
 }: {
   tools: LanguageModelV3CallOptions['tools'];
   toolChoice: LanguageModelV3CallOptions['toolChoice'] | undefined;
@@ -42,6 +43,12 @@ export async function prepareTools({
    * Whether the model supports strict mode on tool definitions.
    */
   supportsStrictTools: boolean;
+
+  /**
+   * Default for `eager_input_streaming` on function tools that do not set
+   * it explicitly. Driven by the model-level `toolStreaming` option.
+   */
+  defaultEagerInputStreaming?: boolean;
 }): Promise<{
   tools: Array<AnthropicTool> | undefined;
   toolChoice: AnthropicToolChoice | undefined;
@@ -73,8 +80,10 @@ export async function prepareTools({
         const anthropicOptions = tool.providerOptions?.anthropic as
           | AnthropicToolOptions
           | undefined;
-        // eager_input_streaming is only supported on custom (function) tools
-        const eagerInputStreaming = anthropicOptions?.eagerInputStreaming;
+        // eager_input_streaming is only supported on custom (function) tools.
+        // Fall back to the model-level default when the tool doesn't set it.
+        const eagerInputStreaming =
+          anthropicOptions?.eagerInputStreaming ?? defaultEagerInputStreaming;
         const deferLoading = anthropicOptions?.deferLoading;
         const allowedCallers = anthropicOptions?.allowedCallers;