fix(max-tokens): anthropic models streaming vs non-streaming (#2999)

* fix(max-tokens): anthropic models streaming vs non-streaming

* default max inc

* update the docs

Author: icecrasher321

apps/docs/content/docs/en/blocks/agent.mdx

@@ -56,6 +56,10 @@ Controls response randomness and creativity:
 - **Medium (0.3-0.7)**: Balanced creativity and focus. Good for general use.
 - **High (0.7-2.0)**: Creative and varied. Ideal for brainstorming and content generation.
 
+### Max Output Tokens
+
+Controls the maximum length of the model's response. For Anthropic models, Sim uses reliable defaults: streaming executions use the model's full capacity (e.g. 64,000 tokens for Claude 4.5), while non-streaming executions default to 8,192 to avoid timeout issues. For long-form content generation via API, explicitly set a higher value.
+
 ### API Key
 
 Your API key for the selected LLM provider. This is securely stored and used for authentication.

apps/sim/blocks/blocks/agent.ts

@@ -513,6 +513,12 @@ Return ONLY the JSON array.`,
         })(),
       }),
     },
+    {
+      id: 'maxTokens',
+      title: 'Max Output Tokens',
+      type: 'short-input',
+      placeholder: 'Enter max tokens (e.g., 4096)...',
+    },
     {
       id: 'responseFormat',
       title: 'Response Format',
@@ -754,6 +760,7 @@ Example 3 (Array Input):
       },
     },
     temperature: { type: 'number', description: 'Response randomness level' },
+    maxTokens: { type: 'number', description: 'Maximum number of tokens in the response' },
     reasoningEffort: { type: 'string', description: 'Reasoning effort level for GPT-5 models' },
     verbosity: { type: 'string', description: 'Verbosity level for GPT-5 models' },
     thinkingLevel: { type: 'string', description: 'Thinking level for Gemini 3 models' },

apps/sim/providers/anthropic/index.ts

@@ -9,6 +9,7 @@ import {
   generateToolUseId,
 } from '@/providers/anthropic/utils'
 import {
+  getMaxOutputTokensForModel,
   getProviderDefaultModel,
   getProviderModels,
   supportsNativeStructuredOutputs,
@@ -178,7 +179,9 @@ export const anthropicProvider: ProviderConfig = {
       model: request.model,
       messages,
       system: systemPrompt,
-      max_tokens: Number.parseInt(String(request.maxTokens)) || 1024,
+      max_tokens:
+        Number.parseInt(String(request.maxTokens)) ||
+        getMaxOutputTokensForModel(request.model, request.stream ?? false),
       temperature: Number.parseFloat(String(request.temperature ?? 0.7)),
     }
 

apps/sim/providers/bedrock/index.ts

@@ -20,7 +20,11 @@ import {
   generateToolUseId,
   getBedrockInferenceProfileId,
 } from '@/providers/bedrock/utils'
-import { getProviderDefaultModel, getProviderModels } from '@/providers/models'
+import {
+  getMaxOutputTokensForModel,
+  getProviderDefaultModel,
+  getProviderModels,
+} from '@/providers/models'
 import type {
   ProviderConfig,
   ProviderRequest,
@@ -259,7 +263,9 @@ export const bedrockProvider: ProviderConfig = {
 
     const inferenceConfig = {
       temperature: Number.parseFloat(String(request.temperature ?? 0.7)),
-      maxTokens: Number.parseInt(String(request.maxTokens)) || 4096,
+      maxTokens:
+        Number.parseInt(String(request.maxTokens)) ||
+        getMaxOutputTokensForModel(request.model, request.stream ?? false),
     }
 
     const shouldStreamToolCalls = request.streamToolCalls ?? false

apps/sim/providers/models.ts

@@ -34,6 +34,12 @@ export interface ModelCapabilities {
   toolUsageControl?: boolean
   computerUse?: boolean
   nativeStructuredOutputs?: boolean
+  maxOutputTokens?: {
+    /** Maximum tokens for streaming requests */
+    max: number
+    /** Safe default for non-streaming requests (to avoid timeout issues) */
+    default: number
+  }
   reasoningEffort?: {
     values: string[]
   }
@@ -613,6 +619,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -627,6 +634,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -640,6 +648,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -654,6 +663,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -668,6 +678,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -681,6 +692,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -695,6 +707,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           computerUse: true,
+          maxOutputTokens: { max: 8192, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -709,6 +722,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           computerUse: true,
+          maxOutputTokens: { max: 8192, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -1655,6 +1669,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -1668,6 +1683,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -1681,6 +1697,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -1694,6 +1711,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -2333,3 +2351,31 @@ export function getThinkingLevelsForModel(modelId: string): string[] | null {
   const capability = getThinkingCapability(modelId)
   return capability?.levels ?? null
 }
+
+/**
+ * Get the max output tokens for a specific model
+ * Returns the model's max capacity for streaming requests,
+ * or the model's safe default for non-streaming requests to avoid timeout issues.
+ *
+ * @param modelId - The model ID
+ * @param streaming - Whether the request is streaming (default: false)
+ */
+export function getMaxOutputTokensForModel(modelId: string, streaming = false): number {
+  const normalizedModelId = modelId.toLowerCase()
+  const STANDARD_MAX_OUTPUT_TOKENS = 4096
+
+  for (const provider of Object.values(PROVIDER_DEFINITIONS)) {
+    for (const model of provider.models) {
+      const baseModelId = model.id.toLowerCase()
+      if (normalizedModelId === baseModelId || normalizedModelId.startsWith(`${baseModelId}-`)) {
+        const outputTokens = model.capabilities.maxOutputTokens
+        if (outputTokens) {
+          return streaming ? outputTokens.max : outputTokens.default
+        }
+        return STANDARD_MAX_OUTPUT_TOKENS
+      }
+    }
+  }
+
+  return STANDARD_MAX_OUTPUT_TOKENS
+}

apps/sim/providers/utils.ts

@@ -8,6 +8,7 @@ import {
   getComputerUseModels,
   getEmbeddingModelPricing,
   getHostedModels as getHostedModelsFromDefinitions,
+  getMaxOutputTokensForModel as getMaxOutputTokensForModelFromDefinitions,
   getMaxTemperature as getMaxTempFromDefinitions,
   getModelPricing as getModelPricingFromDefinitions,
   getModelsWithReasoningEffort,
@@ -992,6 +993,18 @@ export function getThinkingLevelsForModel(model: string): string[] | null {
   return getThinkingLevelsForModelFromDefinitions(model)
 }
 
+/**
+ * Get max output tokens for a specific model
+ * Returns the model's maxOutputTokens capability for streaming requests,
+ * or a conservative default (8192) for non-streaming requests to avoid timeout issues.
+ *
+ * @param model - The model ID
+ * @param streaming - Whether the request is streaming (default: false)
+ */
+export function getMaxOutputTokensForModel(model: string, streaming = false): number {
+  return getMaxOutputTokensForModelFromDefinitions(model, streaming)
+}
+
 /**
  * Prepare tool execution parameters, separating tool parameters from system parameters
  */

apps/sim/tools/browser_use/run_task.ts

@@ -5,7 +5,7 @@ import type { ToolConfig, ToolResponse } from '@/tools/types'
 const logger = createLogger('BrowserUseTool')
 
 const POLL_INTERVAL_MS = 5000
-const MAX_POLL_TIME_MS = 180000
+const MAX_POLL_TIME_MS = 600000 // 10 minutes
 const MAX_CONSECUTIVE_ERRORS = 3
 
 async function createSessionWithProfile(