@@ -158,20 +158,21 @@ Live tool-call chips stream for **OpenAI, Anthropic, Azure Anthropic, Google, Ve
 
 | Provider | Streamed thinking | Models |
 |----------|-------------------|--------|
-| OpenAI | Summaries only — Requires OpenAI organization verification; falls back to no summaries. | `gpt-6-astra`, `gpt-6-sol`, `gpt-6-luna`, `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-5.6-luna`, `gpt-5.5-pro`, `gpt-5.5`, `gpt-5.4-pro`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.4-nano`, `gpt-5.3-codex`, `gpt-5.2-pro`, `gpt-5.2`, `gpt-5.1`, `gpt-5-pro`, `gpt-5`, `gpt-5-mini`, `gpt-5-nano`, `o4-mini`, `o3`, `o3-mini`, `o1` |
-| Anthropic | Summaries only — These generations omit full thinking; Sim requests summarized thinking on streaming runs. | `claude-fable-5-1`, `claude-fable-5`, `claude-sonnet-5-5`, `claude-sonnet-5`, `claude-opus-5-5`, `claude-opus-5`, `claude-opus-4-8`, `claude-opus-4-7`, `claude-opus-4-6`, `claude-sonnet-4-6`, `claude-opus-4-5`, `claude-sonnet-4-5`, `claude-haiku-4-5` |
-| Azure OpenAI | Summaries only — Requires OpenAI organization verification; falls back to no summaries. | `azure/gpt-6-astra`, `azure/gpt-5.6-sol`, `azure/gpt-5.6-terra`, `azure/gpt-5.6-luna`, `azure/gpt-5.5`, `azure/gpt-5.4-pro`, `azure/gpt-5.4`, `azure/gpt-5.4-mini`, `azure/gpt-5.4-nano`, `azure/gpt-5.2`, `azure/gpt-5.1`, `azure/gpt-5.1-codex`, `azure/gpt-5`, `azure/gpt-5-mini`, `azure/gpt-5-nano`, `azure/o3`, `azure/o4-mini` |
-| Azure Anthropic | Summaries only — These generations omit full thinking; Sim requests summarized thinking on streaming runs. | `azure-anthropic/claude-fable-5-1`, `azure-anthropic/claude-opus-5`, `azure-anthropic/claude-opus-4-8`, `azure-anthropic/claude-opus-4-7`, `azure-anthropic/claude-opus-4-6`, `azure-anthropic/claude-opus-4-5`, `azure-anthropic/claude-sonnet-5`, `azure-anthropic/claude-sonnet-4-6`, `azure-anthropic/claude-sonnet-4-5`, `azure-anthropic/claude-opus-4-1`, `azure-anthropic/claude-haiku-4-5` |
+| OpenAI | Summaries only — Requires OpenAI organization verification; falls back to no summaries. | `gpt-6-astra`, `gpt-6.1-sol`, `gpt-6-sol`, `gpt-6-luna`, `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-5.6-luna`, `gpt-5.5-pro`, `gpt-5.5`, `gpt-5.4-pro`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.4-nano`, `gpt-5.3-codex`, `gpt-5.2-pro`, `gpt-5.2`, `gpt-5.1`, `gpt-5-pro`, `gpt-5`, `gpt-5-mini`, `gpt-5-nano`, `o4-mini`, `o3`, `o3-mini`, `o1` |
+| Anthropic | Summaries only — These generations omit full thinking; Sim requests summarized thinking on streaming runs. | `claude-fable-5-1`, `claude-fable-5`, `claude-sonnet-5-5`, `claude-sonnet-5`, `claude-opus-5-5`, `claude-opus-5`, `claude-opus-4-8`, `claude-opus-4-7`, `claude-opus-4-6`, `claude-sonnet-4-6`, `claude-opus-4-5`, `claude-sonnet-4-5`, `claude-haiku-5-5`, `claude-haiku-4-5` |
+| Azure OpenAI | Summaries only — Requires OpenAI organization verification; falls back to no summaries. | `azure/gpt-6.1-sol`, `azure/gpt-6-sol`, `azure/gpt-6-luna`, `azure/gpt-6-astra`, `azure/gpt-5.6-sol`, `azure/gpt-5.6-terra`, `azure/gpt-5.6-luna`, `azure/gpt-5.5`, `azure/gpt-5.4-pro`, `azure/gpt-5.4`, `azure/gpt-5.4-mini`, `azure/gpt-5.4-nano`, `azure/gpt-5.2`, `azure/gpt-5.1`, `azure/gpt-5.1-codex`, `azure/gpt-5`, `azure/gpt-5-mini`, `azure/gpt-5-nano`, `azure/o3`, `azure/o4-mini` |
+| Azure Anthropic | Summaries only — These generations omit full thinking; Sim requests summarized thinking on streaming runs. | `azure-anthropic/claude-fable-5-1`, `azure-anthropic/claude-opus-5-5`, `azure-anthropic/claude-sonnet-5-5`, `azure-anthropic/claude-haiku-5-5`, `azure-anthropic/claude-opus-5`, `azure-anthropic/claude-opus-4-8`, `azure-anthropic/claude-opus-4-7`, `azure-anthropic/claude-opus-4-6`, `azure-anthropic/claude-opus-4-5`, `azure-anthropic/claude-sonnet-5`, `azure-anthropic/claude-sonnet-4-6`, `azure-anthropic/claude-sonnet-4-5`, `azure-anthropic/claude-opus-4-1`, `azure-anthropic/claude-haiku-4-5` |
 | Google | Summaries only | `gemini-3.8-flash`, `gemini-3.7-flash`, `gemini-3.6-flash`, `gemini-3.5-flash-lite`, `gemini-3.5-flash`, `gemini-3.1-pro-preview`, `gemini-3.1-flash-lite`, `gemini-3-flash-preview`, `gemini-2.5-pro`, `gemini-2.5-flash`, `gemini-2.5-flash-lite` |
 | Vertex AI | Summaries only | `vertex/gemini-3.8-flash`, `vertex/gemini-3.7-flash`, `vertex/gemini-3.6-flash`, `vertex/gemini-3.5-flash-lite`, `vertex/gemini-3.5-flash`, `vertex/gemini-3.1-pro-preview`, `vertex/gemini-3.1-flash-lite`, `vertex/gemini-3-flash-preview`, `vertex/gemini-2.5-pro`, `vertex/gemini-2.5-flash`, `vertex/gemini-2.5-flash-lite` |
 | DeepSeek | Full thinking deltas | `deepseek-v4-pro`, `deepseek-flash`, `deepseek-v4-flash`, `deepseek-reasoner` |
 | xAI | Full thinking deltas | `grok-4.6`, `grok-4.5`, `grok-4.3`, `grok-4.20-multi-agent-0309` |
+| xAI | Summaries only | `grok-4.7` |
 | Cerebras | Full thinking deltas | `cerebras/qwen-3.8-27b` |
 | Groq | Full thinking deltas | `groq/openai/gpt-oss-120b`, `groq/openai/gpt-oss-20b`, `groq/openai/gpt-oss-safeguard-20b`, `groq/qwen/qwen3.8-27b`, `groq/qwen/qwen3.6-27b` |
 | Meta | Not streamed | `muse-spark-1.3`, `muse-spark-1.1` |
 | Kimi | Full thinking deltas | `kimi-k3`, `kimi-k2.6` |
-| Kie | Summaries only | `kie/claude-opus-5-5`, `kie/claude-opus-5`, `kie/claude-fable-5`, `kie/claude-sonnet-5`, `kie/claude-opus-4-8`, `kie/claude-opus-4-7`, `kie/claude-opus-4-6`, `kie/claude-sonnet-4-6`, `kie/claude-opus-4-5`, `kie/claude-sonnet-4-5`, `kie/claude-haiku-4-5`, `kie/gpt-6-astra`, `kie/gpt-6-sol`, `kie/gpt-6-luna`, `kie/gpt-5-6-sol`, `kie/gpt-5-6-terra`, `kie/gpt-5-6-luna`, `kie/gpt-5-5`, `kie/grok-4-6`, `kie/grok-4-5`, `kie/grok-4-3`, `kie/kimi-k3` |
-| Z.ai | Full thinking deltas | `glm-5.3`, `glm-5.3-flash`, `glm-5.2`, `glm-5.1`, `glm-5`, `glm-5-turbo`, `glm-4.7`, `glm-4.6`, `glm-4.5`, `glm-4.5-air` |
+| Kie | Summaries only | `kie/claude-sonnet-5-5`, `kie/claude-opus-5-5`, `kie/claude-opus-5`, `kie/claude-fable-5`, `kie/claude-sonnet-5`, `kie/claude-opus-4-8`, `kie/claude-opus-4-7`, `kie/claude-opus-4-6`, `kie/claude-sonnet-4-6`, `kie/claude-opus-4-5`, `kie/claude-sonnet-4-5`, `kie/claude-haiku-4-5`, `kie/gpt-6-1-sol`, `kie/gpt-6-astra`, `kie/gpt-6-sol`, `kie/gpt-6-luna`, `kie/gpt-5-6-sol`, `kie/gpt-5-6-terra`, `kie/gpt-5-6-luna`, `kie/gpt-5-5`, `kie/grok-4-6`, `kie/grok-4-5`, `kie/grok-4-3`, `kie/kimi-k3` |
+| Z.ai | Full thinking deltas | `glm-5.3`, `glm-5.3-flashx`, `glm-5.3-flash`, `glm-5.2`, `glm-5.1`, `glm-5`, `glm-5-turbo`, `glm-4.7`, `glm-4.6`, `glm-4.5`, `glm-4.5-air` |
 
 {/* agent-stream-capabilities:end */}
 
 
@@ -747,14 +747,10 @@ describe('streaming', () => {
   })
 
   /**
-   * Both tool loops rebuild the request for every turn after a tool call, so
-   * the `none` mapping must hold past the first request. Claude Sonnet 5.5
-   * rejects `thinking.type: "disabled"` and names `between_tools` (which takes
-   * no effort or display field) as its lowest setting; a later turn without it
-   * would silently run adaptive thinking. Every other model keeps `none` as
-   * "send no thinking config".
+   * Both tool loops rebuild requests after tool calls. Explicit none modes
+   * must survive later turns or default-thinking models resume thinking.
    */
-  describe('executeAnthropicProviderRequest none thinking level across tool turns', () => {
+  describe('Anthropic tool exchanges', () => {
     const lookupTool = {
       id: 'lookup',
       name: 'lookup',
@@ -767,24 +763,31 @@ describe('streaming', () => {
       message([{ type: 'text', text: 'done' }], 'end_turn'),
     ]
 
-    /** Runs a tool exchange and returns every request body sent over the SDK boundary. */
-    async function runToolExchange(model: string, streaming: boolean) {
+    /** Runs a tool exchange and returns the SDK request bodies and settled output. */
+    async function runToolExchange(
+      model: string,
+      streaming: boolean,
+      thinkingLevel = 'none',
+      forceTool = false,
+      responseTurns = turns
+    ) {
       mockExecuteTool.mockResolvedValue({ success: true, output: { value: 'tool result' } })
       const sent: Anthropic.Messages.MessageCreateParams[] = []
       const nextTurn = (payload: Anthropic.Messages.MessageCreateParams) => {
         sent.push(payload)
-        return turns[sent.length - 1]
+        return responseTurns[sent.length - 1]
       }
       const result = await executeAnthropicProviderRequest(
         {
           model,
           apiKey: 'test-key',
           stream: streaming,
-          maxTokens: 1024,
-          thinkingLevel: 'none',
+          maxTokens: 4096,
+          thinkingLevel,
+          temperature: 0.5,
           agentEvents: true,
           messages: [{ role: 'user', content: 'Look this up' }],
-          tools: [lookupTool],
+          tools: [{ ...lookupTool, usageControl: forceTool ? 'force' : 'auto' }],
         },
         {
           providerId: 'anthropic',
@@ -802,13 +805,18 @@ describe('streaming', () => {
         }
       )
       if (streaming) await collectEvents(result as StreamingExecution)
-      return sent
+      return {
+        sent,
+        output: streaming
+          ? (result as StreamingExecution).execution.output
+          : (result as ProviderResponse),
+      }
     }
 
     it.each([false, true])(
       'sends bare between_tools on every turn (streaming: %s)',
       async (streaming) => {
-        const sent = await runToolExchange('claude-sonnet-5-5', streaming)
+        const { sent } = await runToolExchange('claude-sonnet-5-5', streaming)
         expect(sent).toHaveLength(2)
         for (const payload of sent) {
           expect(payload.thinking).toEqual({ type: 'between_tools' })
@@ -820,11 +828,81 @@ describe('streaming', () => {
     it.each(['claude-sonnet-5', 'claude-opus-5-5'])(
       'sends no thinking config on %s',
       async (model) => {
-        const sent = await runToolExchange(model, false)
+        const { sent } = await runToolExchange(model, false)
         expect(sent).toHaveLength(2)
         for (const payload of sent) expect(payload.thinking).toBeUndefined()
       }
     )
+
+    it.each([false, true])(
+      'disables Haiku 5.5 default thinking on every tool turn (streaming: %s)',
+      async (streaming) => {
+        const { sent } = await runToolExchange('claude-haiku-5-5', streaming)
+        expect(sent).toHaveLength(2)
+        for (const payload of sent) {
+          expect(payload.thinking).toEqual({ type: 'disabled' })
+          expect(payload.output_config).toBeUndefined()
+          expect(payload.temperature).toBeUndefined()
+        }
+      }
+    )
+
+    it.each([false, true])(
+      'uses adaptive Haiku 5.5 thinking and releases the executed forced tool (streaming: %s)',
+      async (streaming) => {
+        const { sent } = await runToolExchange('claude-haiku-5-5', streaming, 'max', true)
+        expect(sent).toHaveLength(2)
+        for (const payload of sent) {
+          expect(payload.thinking).toEqual({ type: 'adaptive', display: 'summarized' })
+          expect(payload.output_config).toEqual({ effort: 'max' })
+          expect(payload.temperature).toBeUndefined()
+        }
+        expect(sent[0].tool_choice).toEqual({ type: 'tool', name: 'lookup' })
+        expect(sent[1].tool_choice).toBeUndefined()
+      }
+    )
+
+    it.each([false, true])(
+      'prices Haiku prompts below the long-context tier separately (streaming: %s)',
+      async (streaming) => {
+        const responseTurns = turns.map((turn) => ({
+          ...turn,
+          usage: { input_tokens: 60000, output_tokens: 2000 },
+        }))
+        const { output } = await runToolExchange(
+          'claude-haiku-5-5',
+          streaming,
+          'none',
+          false,
+          responseTurns
+        )
+        expect(output.tokens).toMatchObject({ input: 120000, output: 4000 })
+        expect(output.cost).toMatchObject({ input: 0.012, output: 0.002, total: 0.014 })
+      }
+    )
+
+    it.each([false, true])(
+      'prices mixed Haiku prompt tiers using each turn including cache reads (streaming: %s)',
+      async (streaming) => {
+        const responseTurns = turns.map((turn, index) => ({
+          ...turn,
+          usage: {
+            input_tokens: index === 0 ? 60000 : 10000,
+            output_tokens: 2000,
+            cache_read_input_tokens: index === 0 ? 10000 : 100000,
+          },
+        }))
+        const { output } = await runToolExchange(
+          'claude-haiku-5-5',
+          streaming,
+          'none',
+          false,
+          responseTurns
+        )
+        expect(output.tokens).toMatchObject({ input: 70000, output: 4000, cacheRead: 110000 })
+        expect(output.cost).toMatchObject({ input: 0.0161, output: 0.006, total: 0.0221 })
+      }
+    )
   })
 })
 
 
@@ -146,6 +146,7 @@ const ANTHROPIC_THINKING_OUTPUT_HEADROOM = 4096
  * Fable 5, Fable 5.1, and Opus 5.5 support ONLY adaptive thinking (always on; type: "disabled" is rejected).
  * Sonnet 5.5 is adaptive by default and rejects type: "disabled"; its lowest setting is "between_tools".
  * Sonnet 5 supports ONLY adaptive thinking (manual budget_tokens returns a 400 error).
+ * Haiku 5.5 supports adaptive thinking and rejects manual budget_tokens.
  * Opus 5, Opus 4.8, and Opus 4.7 support ONLY adaptive thinking (no extended thinking / budget_tokens).
  * Opus 4.6 and Sonnet 4.6 support both extended and adaptive thinking — use adaptive.
  * Opus 4.5 supports effort but NOT adaptive thinking — it uses budget_tokens with type: "enabled".
@@ -155,6 +156,7 @@ function supportsAdaptiveThinking(modelId: string): boolean {
   return (
     normalizedModel.includes('fable-5') ||
     normalizedModel.includes('sonnet-5') ||
+    normalizedModel.includes('haiku-5-5') ||
     normalizedModel.includes('opus-5') ||
     normalizedModel.includes('opus-4-8') ||
     normalizedModel.includes('opus-4.8') ||
@@ -175,9 +177,9 @@ function supportsAdaptiveThinking(modelId: string): boolean {
  * - Other models: Uses budget_tokens-based extended thinking
  *
  * The `none` level returns null (send no thinking config) unless the model
- * declares `capabilities.thinking.noneMode`: Sonnet 5.5 rejects
- * `type: "disabled"`, so `none` becomes `type: "between_tools"`, which turns
- * off up-front thinking and takes no effort or display field.
+ * declares `capabilities.thinking.noneMode`: Sonnet 5.5 uses `between_tools`,
+ * while Haiku 5.5 requires `disabled` to turn off its default thinking.
+ * Neither mode takes an effort or display field.
  *
  * The newest Claude generations default `thinking.display` to `omitted`
  * (empty thinking blocks, no thinking deltas). Their registry entries mark
@@ -201,6 +203,7 @@ export function buildThinkingConfig(
   }
 
   if (thinkingLevel === 'none') {
+    if (capability.noneMode === 'disabled') return { thinking: { type: 'disabled' } }
     if (capability.noneMode !== 'between_tools') return null
     return {
       // double-cast-allowed: @anthropic-ai/sdk 0.115 predates the between_tools thinking type (typed from 0.129)
@@ -221,7 +224,9 @@ export function buildThinkingConfig(
         ...(requestSummarizedDisplay ? { display: 'summarized' as const } : {}),
       },
       // Levels are validated against the model's capability list above.
-      outputConfig: { effort: thinkingLevel as Anthropic.Messages.OutputConfig['effort'] },
+      ...(thinkingLevel !== 'enabled' && {
+        outputConfig: { effort: thinkingLevel as Anthropic.Messages.OutputConfig['effort'] },
+      }),
     }
   }
 
@@ -424,7 +429,7 @@ export async function executeAnthropicProviderRequest(
       payload.temperature = undefined
 
       if (request.thinkingLevel === 'none') {
-        logger.info(`Using between_tools thinking for model: ${modelId}`)
+        logger.info(`Using ${thinkingConfig.thinking.type} thinking for model: ${modelId}`)
       } else {
         const isAdaptive = thinkingConfig.thinking.type === 'adaptive'
         logger.info(
@@ -440,10 +445,11 @@ export async function executeAnthropicProviderRequest(
 
   if (anthropicTools?.length) {
     payload.tools = anthropicTools
-    // Per Anthropic docs: forced tool_choice (type: "tool" or "any") is incompatible with
-    // thinking. Only auto and none are supported when thinking is enabled.
-    if (payload.thinking) {
-      // Per Anthropic docs: only 'auto' (default) and 'none' work with thinking.
+    const restrictForcedTools =
+      payload.thinking &&
+      payload.thinking.type !== 'disabled' &&
+      !getThinkingCapability(request.model)?.forcedToolUse
+    if (restrictForcedTools) {
       if (toolChoice === 'none') {
         payload.tool_choice = { type: 'none' }
       }
@@ -871,15 +877,7 @@ export async function executeAnthropicProviderRequest(
           messages: currentMessages,
         }
 
-        // Per Anthropic docs: forced tool_choice is incompatible with thinking.
-        // Only auto and none are supported when thinking is enabled.
-        const thinkingEnabled = !!payload.thinking
-        if (
-          !thinkingEnabled &&
-          typeof originalToolChoice === 'object' &&
-          hasUsedForcedTool &&
-          forcedTools.length > 0
-        ) {
+        if (typeof originalToolChoice === 'object' && hasUsedForcedTool && forcedTools.length > 0) {
           const remainingTools = forcedTools.filter((tool) => !usedForcedTools.includes(tool))
 
           if (remainingTools.length > 0) {
@@ -892,11 +890,7 @@ export async function executeAnthropicProviderRequest(
             nextPayload.tool_choice = undefined
             logger.info('All forced tools have been used, removing tool_choice parameter')
           }
-        } else if (
-          !thinkingEnabled &&
-          hasUsedForcedTool &&
-          typeof originalToolChoice === 'object'
-        ) {
+        } else if (hasUsedForcedTool && typeof originalToolChoice === 'object') {
           nextPayload.tool_choice = undefined
           logger.info(
             'Removing tool_choice parameter for subsequent requests after forced tool was used'
 
@@ -180,11 +180,8 @@ export function createAnthropicStreamingToolLoopStream(
             turnPayload.tool_choice = { type: 'none' }
           }
 
-          // Forced tool_choice vs thinking — same rules as silent loop.
-          const thinkingEnabled = !!payload.thinking
           if (
             !finalSynthesis &&
-            !thinkingEnabled &&
             typeof originalToolChoice === 'object' &&
             hasUsedForcedTool &&
             forcedToolNames.length > 0
@@ -197,7 +194,6 @@ export function createAnthropicStreamingToolLoopStream(
             }
           } else if (
             !finalSynthesis &&
-            !thinkingEnabled &&
             hasUsedForcedTool &&
             typeof originalToolChoice === 'object'
           ) {