smoltalk 0.15.3 → 0.16.0

package/dist/models.d.ts

@@ -113,6 +113,7 @@ export type EmbeddingsModel = {
     type: "embeddings";
     modelName: string;
     provider: string;
+    disabled?: boolean;
     tokenCost?: number;
 };
 /**
@@ -133,10 +134,17 @@ export type ModelType = SpeechToTextModel | TextToSpeechModel | TextModel | Embe
 export declare const speechToTextModels: readonly [{
     readonly type: "speech-to-text";
     readonly modelName: "whisper-1";
+    readonly description: "Scheduled shutdown February 26, 2027 (announced 2026-08-26); use gpt-transcribe instead.";
     readonly perMinuteCost: 0.006;
     readonly provider: "openai";
     readonly supportedMimeTypes: readonly ["audio/flac", "audio/mpeg", "audio/mp4", "audio/m4a", "audio/ogg", "audio/wav", "audio/webm"];
     readonly maxBytes: number;
+}, {
+    readonly type: "speech-to-text";
+    readonly modelName: "gpt-transcribe";
+    readonly description: "OpenAI's transcription model on v1/audio/transcriptions; the named replacement for whisper-1.";
+    readonly perMinuteCost: 0.0045;
+    readonly provider: "openai";
 }, {
     readonly type: "speech-to-text";
     readonly modelName: "whisper-large-v3";
@@ -190,6 +198,22 @@ export declare const textToSpeechModels: readonly [{
     readonly perCharacterCost: 0.00004;
     readonly maxInputChars: 200;
     readonly formats: readonly ["wav"];
+}, {
+    readonly type: "text-to-speech";
+    readonly modelName: "gemini-3.8-flash-tts";
+    readonly provider: "google";
+    readonly description: "Gemini 3.8 Flash TTS (GA 2026-09-22). Introductory pricing through December 31, 2026; $1.00 input / $18.00 audio output per 1M from January 1, 2027.";
+    readonly inputTokenCost: 0.5;
+    readonly outputAudioTokenCost: 9;
+    readonly formats: readonly ["pcm", "wav"];
+}, {
+    readonly type: "text-to-speech";
+    readonly modelName: "gemini-3.8-flash-lite-tts";
+    readonly provider: "google";
+    readonly description: "Gemini 3.8 Flash-Lite TTS (GA 2026-09-22). Introductory pricing through December 31, 2026; $1.00 input / $12.00 audio output per 1M from January 1, 2027.";
+    readonly inputTokenCost: 0.5;
+    readonly outputAudioTokenCost: 6;
+    readonly formats: readonly ["pcm", "wav"];
 }, {
     readonly type: "text-to-speech";
     readonly modelName: "gemini-3.1-flash-tts-preview";
@@ -221,7 +245,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 0.15;
     readonly cachedInputTokenCost: 0.075;
     readonly outputTokenCost: 0.6;
-    readonly outputTokensPerSecond: 65;
+    readonly outputTokensPerSecond: 110;
     readonly modalities: {
         readonly input: readonly ["text", "image", "pdf"];
         readonly output: readonly ["text"];
@@ -259,7 +283,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "o3";
-    readonly description: "o3 is a reasoning model that sets a new standard for math, science, coding, visual reasoning tasks, and technical writing. Part of the o-series of reasoning models. Knowledge cutoff: June 2024.";
+    readonly description: "o3 is a reasoning model that sets a new standard for math, science, coding, visual reasoning tasks, and technical writing. Part of the o-series of reasoning models. Knowledge cutoff: June 2024. Scheduled shutdown December 11, 2026 (announced 2026-06-11); migrate to the gpt-5.6 family.";
     readonly maxInputTokens: 200000;
     readonly maxOutputTokens: 100000;
     readonly inputTokenCost: 2;
@@ -288,7 +312,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "o3-mini";
-    readonly description: "o3-mini is our most recent small reasoning model, providing high intelligence at the same cost and latency targets of o1-mini. o3-mini also supports key developer features, like Structured Outputs, function calling, Batch API, and more. Like other models in the o-series, it is designed to excel at science, math, and coding tasks. Knowledge cutoff: June 2024.";
+    readonly description: "o3-mini is our most recent small reasoning model, providing high intelligence at the same cost and latency targets of o1-mini. o3-mini also supports key developer features, like Structured Outputs, function calling, Batch API, and more. Like other models in the o-series, it is designed to excel at science, math, and coding tasks. Knowledge cutoff: June 2024. Scheduled shutdown October 23, 2026; migrate to gpt-5.6-sol.";
     readonly maxInputTokens: 200000;
     readonly maxOutputTokens: 100000;
     readonly inputTokenCost: 1.1;
@@ -317,7 +341,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "o4-mini";
-    readonly description: "Latest small o-series model optimized for fast, effective reasoning with exceptional performance in coding and visual tasks. Knowledge cutoff: June 2024.";
+    readonly description: "Latest small o-series model optimized for fast, effective reasoning with exceptional performance in coding and visual tasks. Knowledge cutoff: June 2024. Scheduled shutdown October 23, 2026; migrate to gpt-5.6-terra.";
     readonly maxInputTokens: 200000;
     readonly maxOutputTokens: 100000;
     readonly inputTokenCost: 1.1;
@@ -346,7 +370,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "o3-pro";
-    readonly description: "o3-pro uses more compute for complex reasoning tasks. Available via Responses API only. Requests may take several minutes. Knowledge cutoff: June 2024.";
+    readonly description: "o3-pro uses more compute for complex reasoning tasks. Available via Responses API only. Requests may take several minutes. Knowledge cutoff: June 2024. Scheduled shutdown December 11, 2026 (announced 2026-06-11); migrate to gpt-5.6-sol with pro reasoning.";
     readonly maxInputTokens: 200000;
     readonly maxOutputTokens: 100000;
     readonly inputTokenCost: 20;
@@ -371,7 +395,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "o1";
-    readonly description: "o1 is a reasoning model designed to excel at complex reasoning tasks including science, math, and coding. The knowledge cutoff for o1 models is October, 2023.";
+    readonly description: "o1 is a reasoning model designed to excel at complex reasoning tasks including science, math, and coding. The knowledge cutoff for o1 models is October, 2023. Scheduled shutdown October 23, 2026; migrate to gpt-5.6-sol.";
     readonly maxInputTokens: 200000;
     readonly maxOutputTokens: 100000;
     readonly inputTokenCost: 15;
@@ -470,7 +494,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 2;
     readonly cachedInputTokenCost: 0.5;
     readonly outputTokenCost: 8;
-    readonly outputTokensPerSecond: 105;
+    readonly outputTokensPerSecond: 141;
     readonly modalities: {
         readonly input: readonly ["text", "image", "pdf"];
         readonly output: readonly ["text"];
@@ -492,7 +516,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 0.4;
     readonly cachedInputTokenCost: 0.1;
     readonly outputTokenCost: 1.6;
-    readonly outputTokensPerSecond: 78;
+    readonly outputTokensPerSecond: 142;
     readonly modalities: {
         readonly input: readonly ["text", "image", "pdf"];
         readonly output: readonly ["text"];
@@ -508,7 +532,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gpt-4.1-nano";
-    readonly description: "GPT-4.1 nano is the fastest and most affordable GPT-4.1 variant with 1M token context window. Knowledge cutoff: June 2024.";
+    readonly description: "GPT-4.1 nano is the fastest and most affordable GPT-4.1 variant with 1M token context window. Knowledge cutoff: June 2024. Scheduled shutdown October 23, 2026; migrate to gpt-5.6-luna.";
     readonly maxInputTokens: 1047576;
     readonly maxOutputTokens: 32768;
     readonly inputTokenCost: 0.1;
@@ -530,7 +554,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gpt-5";
-    readonly description: "GPT-5 is a frontier reasoning model with 400K context window. Supports reasoning tokens. Knowledge cutoff: September 2024.";
+    readonly description: "GPT-5 is a frontier reasoning model with 400K context window. Supports reasoning tokens. Knowledge cutoff: September 2024. Scheduled shutdown December 11, 2026 (announced 2026-06-11); migrate to the gpt-5.6 family.";
     readonly maxInputTokens: 400000;
     readonly maxOutputTokens: 128000;
     readonly inputTokenCost: 1.25;
@@ -559,13 +583,13 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gpt-5-mini";
-    readonly description: "GPT-5 mini is a faster, more cost-efficient version of GPT-5 with 400K context window. Knowledge cutoff: May 2024.";
+    readonly description: "GPT-5 mini is a faster, more cost-efficient version of GPT-5 with 400K context window. Knowledge cutoff: May 2024. Scheduled shutdown December 11, 2026 (announced 2026-06-11); migrate to the gpt-5.6 family.";
     readonly maxInputTokens: 400000;
     readonly maxOutputTokens: 128000;
     readonly inputTokenCost: 0.25;
     readonly cachedInputTokenCost: 0.025;
     readonly outputTokenCost: 2;
-    readonly outputTokensPerSecond: 69;
+    readonly outputTokensPerSecond: 127;
     readonly reasoning: {
         readonly levels: readonly ["minimal", "low", "medium", "high"];
         readonly defaultLevel: "medium";
@@ -588,7 +612,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gpt-5-nano";
-    readonly description: "GPT-5 nano is the fastest and most affordable GPT-5 variant with 400K context window. Knowledge cutoff: May 2024.";
+    readonly description: "GPT-5 nano is the fastest and most affordable GPT-5 variant with 400K context window. Knowledge cutoff: May 2024. Scheduled shutdown December 11, 2026 (announced 2026-06-11); migrate to the gpt-5.6 family.";
     readonly maxInputTokens: 400000;
     readonly maxOutputTokens: 128000;
     readonly inputTokenCost: 0.05;
@@ -623,6 +647,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 1.25;
     readonly cachedInputTokenCost: 0.125;
     readonly outputTokenCost: 10;
+    readonly outputTokensPerSecond: 106;
     readonly reasoning: {
         readonly levels: readonly ["none", "low", "medium", "high"];
         readonly defaultLevel: "none";
@@ -651,7 +676,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 1.75;
     readonly cachedInputTokenCost: 0.175;
     readonly outputTokenCost: 14;
-    readonly outputTokensPerSecond: 61;
+    readonly outputTokensPerSecond: 79;
     readonly reasoning: {
         readonly levels: readonly ["none", "low", "medium", "high"];
         readonly defaultLevel: "none";
@@ -705,6 +730,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 2.5;
     readonly cachedInputTokenCost: 0.25;
     readonly outputTokenCost: 15;
+    readonly outputTokensPerSecond: 88;
     readonly longContext: {
         readonly inputTokenCost: 5;
         readonly cachedInputTokenCost: 0.5;
@@ -739,7 +765,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 0.75;
     readonly cachedInputTokenCost: 0.075;
     readonly outputTokenCost: 4.5;
-    readonly outputTokensPerSecond: 180;
+    readonly outputTokensPerSecond: 216;
     readonly reasoning: {
         readonly levels: readonly ["none", "low", "medium", "high", "xhigh"];
         readonly defaultLevel: "none";
@@ -829,7 +855,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 5;
     readonly cachedInputTokenCost: 0.5;
     readonly outputTokenCost: 30;
-    readonly outputTokensPerSecond: 66;
+    readonly outputTokensPerSecond: 86;
     readonly longContext: {
         readonly inputTokenCost: 10;
         readonly cachedInputTokenCost: 1;
@@ -890,13 +916,13 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gpt-6-astra";
-    readonly description: "GPT-6 Astra is OpenAI's most capable model, built for the hardest end-to-end work: complex reasoning, coding, computer use, research, and document creation. 1M context window. Standard pricing for ≤272K input tokens; prompts above that are billed at 2x input/cache and 1.5x output for the whole request. Knowledge cutoff: April 2026.";
+    readonly description: "GPT-6 Astra is OpenAI's most capable model, built for the hardest end-to-end work: complex reasoning, coding, computer use, research, and document creation. 1M context window. Standard pricing for ≤272K input tokens; prompts above that are billed at 2x input/cache and 1.5x output for the whole request. Does not support custom temperature/top_p or logprobs. Knowledge cutoff: April 2026.";
     readonly maxInputTokens: 1050000;
     readonly maxOutputTokens: 128000;
     readonly inputTokenCost: 10;
     readonly cachedInputTokenCost: 1;
     readonly outputTokenCost: 50;
-    readonly outputTokensPerSecond: 69;
+    readonly outputTokensPerSecond: 52;
     readonly longContext: {
         readonly inputTokenCost: 20;
         readonly cachedInputTokenCost: 2;
@@ -915,8 +941,114 @@ export declare const textModels: readonly [{
         readonly output: readonly ["text"];
     };
     readonly knowledge: "2026-04-30";
-    readonly releaseDate: "2026-09-04";
-    readonly lastUpdated: "2026-09-04";
+    readonly releaseDate: "2026-09-03";
+    readonly lastUpdated: "2026-09-03";
+    readonly family: "gpt";
+    readonly openWeights: false;
+    readonly structuredOutput: true;
+    readonly temperatureSupported: false;
+    readonly provider: "openai";
+}, {
+    readonly type: "text";
+    readonly modelName: "gpt-6.1-sol";
+    readonly description: "GPT-6.1 Sol (released 2026-09-29) is the newer Sol model, replacing gpt-6-sol in OpenAI's flagship lineup at the same per-token price; cache reads are 0.05x input ($0.10/1M) instead of 0.1x. 1M context window. Standard pricing for ≤272K input tokens; prompts above that are billed at 2x input/cache and 1.5x output for the whole request. Reasoning cannot be set to none. Cataloged under openai-responses because tool calling requires the Responses API (Chat Completions works only without tools). Knowledge cutoff: April 2026.";
+    readonly maxInputTokens: 1050000;
+    readonly maxOutputTokens: 128000;
+    readonly inputTokenCost: 2;
+    readonly cachedInputTokenCost: 0.1;
+    readonly cacheCreationInputTokenCost: 2.5;
+    readonly outputTokenCost: 10;
+    readonly outputTokensPerSecond: 64;
+    readonly longContext: {
+        readonly inputTokenCost: 4;
+        readonly cachedInputTokenCost: 0.2;
+        readonly outputTokenCost: 15;
+        readonly thresholdTokens: 272000;
+    };
+    readonly reasoning: {
+        readonly levels: readonly ["low", "medium", "high", "xhigh", "max"];
+        readonly defaultLevel: "medium";
+        readonly canDisable: false;
+        readonly outputsThinking: false;
+        readonly outputsSignatures: false;
+    };
+    readonly modalities: {
+        readonly input: readonly ["text", "image"];
+        readonly output: readonly ["text"];
+    };
+    readonly knowledge: "2026-04-30";
+    readonly releaseDate: "2026-09-29";
+    readonly lastUpdated: "2026-09-29";
+    readonly family: "gpt";
+    readonly openWeights: false;
+    readonly structuredOutput: true;
+    readonly temperatureSupported: false;
+    readonly provider: "openai-responses";
+}, {
+    readonly type: "text";
+    readonly modelName: "gpt-6-sol";
+    readonly description: "GPT-6 Sol balances capability and cost in the GPT-6 family. Superseded by gpt-6.1-sol (same price, cheaper cache reads); not deprecated. 1M context window. Standard pricing for ≤272K input tokens; prompts above that are billed at 2x input/cache and 1.5x output for the whole request. Knowledge cutoff: April 2026.";
+    readonly maxInputTokens: 1050000;
+    readonly maxOutputTokens: 128000;
+    readonly inputTokenCost: 2;
+    readonly cachedInputTokenCost: 0.2;
+    readonly outputTokenCost: 10;
+    readonly outputTokensPerSecond: 87;
+    readonly longContext: {
+        readonly inputTokenCost: 4;
+        readonly cachedInputTokenCost: 0.4;
+        readonly outputTokenCost: 15;
+        readonly thresholdTokens: 272000;
+    };
+    readonly reasoning: {
+        readonly levels: readonly ["none", "low", "medium", "high", "xhigh", "max"];
+        readonly defaultLevel: "medium";
+        readonly canDisable: true;
+        readonly outputsThinking: false;
+        readonly outputsSignatures: false;
+    };
+    readonly modalities: {
+        readonly input: readonly ["text", "image"];
+        readonly output: readonly ["text"];
+    };
+    readonly knowledge: "2026-04-20";
+    readonly releaseDate: "2026-09-22";
+    readonly lastUpdated: "2026-09-22";
+    readonly family: "gpt";
+    readonly openWeights: false;
+    readonly structuredOutput: true;
+    readonly temperatureSupported: false;
+    readonly provider: "openai";
+}, {
+    readonly type: "text";
+    readonly modelName: "gpt-6-luna";
+    readonly description: "GPT-6 Luna is the fast, most affordable member of the GPT-6 family. 1M context window. Standard pricing for ≤272K input tokens; prompts above that are billed at 2x input/cache and 1.5x output for the whole request. Knowledge cutoff: May 2026.";
+    readonly maxInputTokens: 1050000;
+    readonly maxOutputTokens: 128000;
+    readonly inputTokenCost: 0.1;
+    readonly cachedInputTokenCost: 0.01;
+    readonly outputTokenCost: 0.5;
+    readonly outputTokensPerSecond: 128;
+    readonly longContext: {
+        readonly inputTokenCost: 0.2;
+        readonly cachedInputTokenCost: 0.02;
+        readonly outputTokenCost: 0.75;
+        readonly thresholdTokens: 272000;
+    };
+    readonly reasoning: {
+        readonly levels: readonly ["none", "low", "medium", "high", "xhigh", "max"];
+        readonly defaultLevel: "medium";
+        readonly canDisable: true;
+        readonly outputsThinking: false;
+        readonly outputsSignatures: false;
+    };
+    readonly modalities: {
+        readonly input: readonly ["text", "image"];
+        readonly output: readonly ["text"];
+    };
+    readonly knowledge: "2026-05-18";
+    readonly releaseDate: "2026-09-22";
+    readonly lastUpdated: "2026-09-22";
     readonly family: "gpt";
     readonly openWeights: false;
     readonly structuredOutput: true;
@@ -1001,7 +1133,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 0.2;
     readonly cachedInputTokenCost: 0.02;
     readonly outputTokenCost: 1.2;
-    readonly outputTokensPerSecond: 165;
+    readonly outputTokensPerSecond: 124;
     readonly longContext: {
         readonly inputTokenCost: 0.4;
         readonly cachedInputTokenCost: 0.04;
@@ -1093,14 +1225,14 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gemini-3.8-flash";
-    readonly description: "Google's most intelligent Flash model (GA September 2, 2026), built on Gemini 3.7 Flash and engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows. 1M context window, 64K output. Introductory pricing $0.75 input / $3.75 output per 1M through December 31, 2026 ($1.50 / $7.50 thereafter). Thinking levels low/medium/high — `minimal` returns an error.";
+    readonly description: "Google's most intelligent Flash model (GA September 2, 2026), built on Gemini 3.7 Flash and engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows. 1M context window, 64K output. Introductory pricing $0.75 input (all modalities, including audio) / $3.75 output per 1M through December 31, 2026 ($1.50 / $7.50 thereafter). Thinking levels low/medium/high — `minimal` returns an error.";
     readonly maxInputTokens: 1048576;
     readonly maxOutputTokens: 65536;
     readonly inputTokenCost: 0.75;
     readonly cachedInputTokenCost: 0.075;
     readonly outputTokenCost: 3.75;
-    readonly outputTokensPerSecond: 329;
-    readonly inputAudioTokenCost: 1.5;
+    readonly outputTokensPerSecond: 249;
+    readonly inputAudioTokenCost: 0.75;
     readonly reasoning: {
         readonly levels: readonly ["low", "medium", "high"];
         readonly defaultLevel: "medium";
@@ -1123,14 +1255,14 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gemini-3.7-flash";
-    readonly description: "Gemini 3.7 Flash (GA August 13, 2026) for complex coding, agentic workflows, and reliable multi-step execution. Superseded by gemini-3.8-flash but fully supported. 1M context window, 64K output. Introductory pricing $0.75 input / $3.75 output per 1M through December 31, 2026 ($1.50 / $7.50 thereafter). Thinking levels low/medium/high — `minimal` returns an error.";
+    readonly description: "Gemini 3.7 Flash (GA August 13, 2026) for complex coding, agentic workflows, and reliable multi-step execution. Superseded by gemini-3.8-flash but fully supported. 1M context window, 64K output. Introductory pricing $0.75 input (all modalities, including audio) / $3.75 output per 1M through December 31, 2026 ($1.50 / $7.50 thereafter). Thinking levels low/medium/high — `minimal` returns an error.";
     readonly maxInputTokens: 1048576;
     readonly maxOutputTokens: 65536;
     readonly inputTokenCost: 0.75;
     readonly cachedInputTokenCost: 0.075;
     readonly outputTokenCost: 3.75;
     readonly outputTokensPerSecond: 272;
-    readonly inputAudioTokenCost: 1.5;
+    readonly inputAudioTokenCost: 0.75;
     readonly reasoning: {
         readonly levels: readonly ["low", "medium", "high"];
         readonly defaultLevel: "medium";
@@ -1153,16 +1285,17 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gemini-3.6-flash";
-    readonly description: "Gemini 3.6 Flash (GA July 21, 2026), balancing speed and multimodal capabilities across general agentic and everyday tasks. Superseded by gemini-3.7-flash / gemini-3.8-flash. 1M context window, 64K output. Introductory pricing $0.75 input / $3.75 output per 1M through December 31, 2026 ($1.50 / $7.50 thereafter). Context caching: $0.075/1M read.";
+    readonly description: "Gemini 3.6 Flash (GA July 21, 2026), balancing speed and multimodal capabilities across general agentic and everyday tasks. Superseded by gemini-3.7-flash / gemini-3.8-flash. 1M context window, 64K output. Introductory pricing $0.75 input (all modalities, including audio) / $3.75 output per 1M through December 31, 2026 ($1.50 / $7.50 thereafter). Context caching: $0.075/1M read.";
     readonly maxInputTokens: 1048576;
     readonly maxOutputTokens: 65536;
     readonly inputTokenCost: 0.75;
     readonly cachedInputTokenCost: 0.075;
     readonly outputTokenCost: 3.75;
-    readonly inputAudioTokenCost: 1.5;
+    readonly outputTokensPerSecond: 185;
+    readonly inputAudioTokenCost: 0.75;
     readonly reasoning: {
         readonly levels: readonly ["minimal", "low", "medium", "high"];
-        readonly defaultLevel: "high";
+        readonly defaultLevel: "medium";
         readonly canDisable: false;
         readonly outputsThinking: true;
         readonly outputsSignatures: true;
@@ -1188,10 +1321,11 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 1.5;
     readonly cachedInputTokenCost: 0.15;
     readonly outputTokenCost: 9;
+    readonly outputTokensPerSecond: 211;
     readonly inputAudioTokenCost: 1.5;
     readonly reasoning: {
         readonly levels: readonly ["minimal", "low", "medium", "high"];
-        readonly defaultLevel: "high";
+        readonly defaultLevel: "medium";
         readonly canDisable: false;
         readonly outputsThinking: true;
         readonly outputsSignatures: true;
@@ -1211,7 +1345,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gemini-3-flash-preview";
-    readonly description: "Gemini 3 Flash preview. Superseded by gemini-3.5-flash. 1M context window and 64K output. Optimized for agentic workflows and coding.";
+    readonly description: "Gemini 3 Flash preview. Superseded by gemini-3.6-flash (Google's named replacement); no shutdown date announced. 1M context window and 64K output. Optimized for agentic workflows and coding.";
     readonly maxInputTokens: 1048576;
     readonly maxOutputTokens: 65536;
     readonly inputTokenCost: 0.5;
@@ -1248,7 +1382,7 @@ export declare const textModels: readonly [{
     readonly cachedInputTokenCost: 0.03;
     readonly outputTokenCost: 2.5;
     readonly outputTokensPerSecond: 347;
-    readonly inputAudioTokenCost: 0.5;
+    readonly inputAudioTokenCost: 0.3;
     readonly reasoning: {
         readonly levels: readonly ["minimal", "low", "medium", "high"];
         readonly defaultLevel: "minimal";
@@ -1271,13 +1405,13 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gemini-3.1-flash-lite";
-    readonly description: "Cost-effective Gemini 3.1 model (GA). Superseded by gemini-3.5-flash-lite. Thinking support, 1M context window, 64K output. 2.5x faster TTFA and 45% faster output than 2.5 Flash.";
+    readonly description: "Cost-effective Gemini 3.1 model (GA). Superseded by gemini-3.5-flash-lite; scheduled shutdown May 7, 2027. Thinking support, 1M context window, 64K output. 2.5x faster TTFA and 45% faster output than 2.5 Flash.";
     readonly maxInputTokens: 1048576;
     readonly maxOutputTokens: 65536;
     readonly inputTokenCost: 0.25;
     readonly cachedInputTokenCost: 0.025;
     readonly outputTokenCost: 1.5;
-    readonly outputTokensPerSecond: 379;
+    readonly outputTokensPerSecond: 267;
     readonly inputAudioTokenCost: 0.5;
     readonly reasoning: {
         readonly levels: readonly ["minimal", "low", "medium", "high"];
@@ -1301,7 +1435,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gemini-3.1-flash-lite-preview";
-    readonly description: "DEPRECATED: Preview version, discontinued July 9, 2026. Use gemini-3.1-flash-lite instead.";
+    readonly description: "DEPRECATED: Preview version, shut down May 25, 2026. Use gemini-3.1-flash-lite instead.";
     readonly maxInputTokens: 1048576;
     readonly maxOutputTokens: 65536;
     readonly inputTokenCost: 0.25;
@@ -1324,13 +1458,13 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gemini-2.5-pro";
-    readonly description: "High-performance Gemini 2.5 model with 1M context window. Adaptive thinking for complex reasoning and coding. Standard pricing for ≤200k tokens ($1.25 input/$10.00 output), higher rates for >200k tokens ($2.50 input/$15.00 output). Batch API: 50% discount.";
+    readonly description: "High-performance Gemini 2.5 model with 1M context window. Adaptive thinking for complex reasoning and coding. Standard pricing for ≤200k tokens ($1.25 input/$10.00 output), higher rates for >200k tokens ($2.50 input/$15.00 output). Batch API: 50% discount. Since September 18, 2026, access is limited to projects that have used it before (not deprecated; served until further notice).";
     readonly maxInputTokens: 1048576;
     readonly maxOutputTokens: 65536;
     readonly inputTokenCost: 1.25;
     readonly cachedInputTokenCost: 0.125;
     readonly outputTokenCost: 10;
-    readonly outputTokensPerSecond: 145;
+    readonly outputTokensPerSecond: 122;
     readonly longContext: {
         readonly inputTokenCost: 2.5;
         readonly cachedInputTokenCost: 0.25;
@@ -1357,7 +1491,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gemini-2.5-flash";
-    readonly description: "Balanced Gemini 2.5 model with excellent performance-to-cost ratio. Lightning-fast with controllable thinking budgets. 1M context window. Context caching available for up to 75% cost reduction.";
+    readonly description: "Balanced Gemini 2.5 model with excellent performance-to-cost ratio. Lightning-fast with controllable thinking budgets. 1M context window. Context caching available for up to 75% cost reduction. Since September 18, 2026, access is limited to projects that have used it before (not deprecated; served until further notice).";
     readonly maxInputTokens: 1048576;
     readonly maxOutputTokens: 65536;
     readonly inputTokenCost: 0.3;
@@ -1387,13 +1521,13 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gemini-2.5-flash-lite";
-    readonly description: "Most cost-effective Gemini 2.5 option for high-throughput applications. 1M context window.";
+    readonly description: "Most cost-effective Gemini 2.5 option for high-throughput applications. 1M context window. Since September 18, 2026, access is limited to projects that have used it before (not deprecated; served until further notice).";
     readonly maxInputTokens: 1048576;
     readonly maxOutputTokens: 65536;
     readonly inputTokenCost: 0.1;
     readonly cachedInputTokenCost: 0.01;
     readonly outputTokenCost: 0.4;
-    readonly outputTokensPerSecond: 400;
+    readonly outputTokensPerSecond: 297;
     readonly inputAudioTokenCost: 0.3;
     readonly reasoning: {
         readonly canDisable: true;
@@ -1501,10 +1635,40 @@ export declare const textModels: readonly [{
     readonly costUnit: "characters";
     readonly disabled: true;
     readonly provider: "google";
+}, {
+    readonly type: "text";
+    readonly modelName: "claude-opus-5-5";
+    readonly description: "Anthropic's recommended default model (released 2026-09-22), superseding claude-opus-5 at a lower price. Cache reads are 0.05x base input ($0.20/MTok) rather than the standard 0.1x. Adaptive thinking is always on (cannot be disabled — `disabled` or `budget_tokens` returns a 400) and defaults to effort `medium`; thinking text is omitted unless requested. Forced tool use (tool_choice any/tool) returns a 400. 1M context window, 128K max output.";
+    readonly maxInputTokens: 1000000;
+    readonly maxOutputTokens: 128000;
+    readonly inputTokenCost: 4;
+    readonly cachedInputTokenCost: 0.2;
+    readonly cacheCreationInputTokenCost: 5;
+    readonly outputTokenCost: 20;
+    readonly outputTokensPerSecond: 92;
+    readonly reasoning: {
+        readonly thinkingStyle: "adaptive";
+        readonly levels: readonly ["low", "medium", "high", "xhigh", "max"];
+        readonly defaultLevel: "medium";
+        readonly canDisable: false;
+        readonly outputsThinking: true;
+        readonly outputsSignatures: true;
+    };
+    readonly modalities: {
+        readonly input: readonly ["text", "image", "pdf"];
+        readonly output: readonly ["text"];
+    };
+    readonly knowledge: "2026-06";
+    readonly releaseDate: "2026-09-22";
+    readonly lastUpdated: "2026-09-22";
+    readonly family: "claude-opus";
+    readonly openWeights: false;
+    readonly temperatureSupported: false;
+    readonly provider: "anthropic";
 }, {
     readonly type: "text";
     readonly modelName: "claude-fable-5-1";
-    readonly description: "Anthropic's most capable widely released model, for the most demanding reasoning and long-horizon agentic work. Successor to claude-fable-5 in the same tier at the same per-token price, but cache reads are 0.025x base input ($0.25/MTok) rather than the standard 0.1x. Thinking is always on (cannot be disabled); the raw chain of thought is never returned. Forced tool use (tool_choice any/tool) returns a 400. Requires 30-day data retention (not available under zero data retention). 1M context window, 128K max output.";
+    readonly description: "Anthropic's most capable widely released model, for the most demanding reasoning and long-horizon agentic work (claude-opus-5-5 is the recommended default for everything else). Successor to claude-fable-5 in the same tier at the same per-token price, but cache reads are 0.025x base input ($0.25/MTok) rather than the standard 0.1x. Thinking is always on (cannot be disabled); the raw chain of thought is never returned. Forced tool use (tool_choice any/tool) returns a 400. Requires 30-day data retention (not available under zero data retention). 1M context window, 128K max output.";
     readonly maxInputTokens: 1000000;
     readonly maxOutputTokens: 128000;
     readonly inputTokenCost: 10;
@@ -1534,7 +1698,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "claude-opus-5";
-    readonly description: "Opus-tier model above Opus 4.8, strongest on deep reasoning, agentic and long-horizon work, at half the cost of claude-fable-5-1. Thinking is on by default; it can be disabled only at effort `high` or below (xhigh/max return a 400). 1M context window, 128K max output.";
+    readonly description: "Legacy Opus-tier model, superseded by claude-opus-5-5 (which is cheaper at $4/$20). Strong on deep reasoning, agentic and long-horizon work. Thinking is on by default; it can be disabled only at effort `high` or below (xhigh/max return a 400). 1M context window, 128K max output.";
     readonly maxInputTokens: 1000000;
     readonly maxOutputTokens: 128000;
     readonly inputTokenCost: 5;
@@ -1594,16 +1758,19 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "claude-opus-4-8";
-    readonly description: "The most capable Claude model for complex reasoning and agentic coding. Same per-token pricing as Opus 4.7 with improved tool-use efficiency (~290 tokens for tool-use system prompt vs 675 on 4.7). 1M context window, 128K max output.";
+    readonly description: "Claude Opus 4.8 — legacy model; prefer claude-opus-5-5. Same per-token pricing as Opus 4.7 with improved tool-use efficiency (~290 tokens for tool-use system prompt vs 675 on 4.7). Adaptive thinking is off by default. 1M context window, 128K max output.";
     readonly maxInputTokens: 1000000;
     readonly maxOutputTokens: 128000;
     readonly inputTokenCost: 5;
     readonly cachedInputTokenCost: 0.5;
     readonly cacheCreationInputTokenCost: 6.25;
     readonly outputTokenCost: 25;
+    readonly outputTokensPerSecond: 61;
     readonly reasoning: {
         readonly thinkingStyle: "adaptive";
-        readonly canDisable: false;
+        readonly levels: readonly ["low", "medium", "high", "xhigh", "max"];
+        readonly defaultLevel: "high";
+        readonly canDisable: true;
         readonly outputsThinking: true;
         readonly outputsSignatures: true;
     };
@@ -1611,6 +1778,7 @@ export declare const textModels: readonly [{
         readonly input: readonly ["text", "image", "pdf"];
         readonly output: readonly ["text"];
     };
+    readonly knowledge: "2026-01";
     readonly releaseDate: "2026-05-28";
     readonly lastUpdated: "2026-05-28";
     readonly family: "claude-opus";
@@ -1627,10 +1795,12 @@ export declare const textModels: readonly [{
     readonly cachedInputTokenCost: 0.5;
     readonly cacheCreationInputTokenCost: 6.25;
     readonly outputTokenCost: 25;
-    readonly outputTokensPerSecond: 72;
+    readonly outputTokensPerSecond: 48;
     readonly reasoning: {
         readonly thinkingStyle: "adaptive";
-        readonly canDisable: false;
+        readonly levels: readonly ["low", "medium", "high", "xhigh", "max"];
+        readonly defaultLevel: "high";
+        readonly canDisable: true;
         readonly outputsThinking: true;
         readonly outputsSignatures: true;
     };
@@ -1648,16 +1818,18 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "claude-opus-4-6";
-    readonly description: "Claude Opus 4.6 — legacy model. Use claude-opus-4-7 instead. 1M context window, 128K max output.";
+    readonly description: "Claude Opus 4.6 — legacy model. Use claude-opus-5-5 instead. 1M context window, 128K max output.";
     readonly maxInputTokens: 1000000;
     readonly maxOutputTokens: 128000;
     readonly inputTokenCost: 5;
     readonly cachedInputTokenCost: 0.5;
     readonly cacheCreationInputTokenCost: 6.25;
     readonly outputTokenCost: 25;
-    readonly outputTokensPerSecond: 53;
+    readonly outputTokensPerSecond: 38;
     readonly reasoning: {
         readonly thinkingStyle: "adaptive";
+        readonly levels: readonly ["low", "medium", "high", "max"];
+        readonly defaultLevel: "high";
         readonly canDisable: true;
         readonly outputsThinking: true;
         readonly outputsSignatures: true;
@@ -1673,10 +1845,39 @@ export declare const textModels: readonly [{
     readonly openWeights: false;
     readonly temperatureSupported: true;
     readonly provider: "anthropic";
+}, {
+    readonly type: "text";
+    readonly modelName: "claude-sonnet-5-5";
+    readonly description: "Current Sonnet-tier model (released 2026-09-28), superseding claude-sonnet-5 at the same price. Adaptive thinking is always on and defaults to effort `high`; `thinking: {type: \"disabled\"}` and `budget_tokens` return a 400. The lowest setting is `thinking: {type: \"between_tools\"}` (low/medium/high effort only), which skips up-front thinking but still emits short thinking blocks between tool calls. Thinking text is omitted unless requested. Forced tool use (tool_choice any/tool) and non-default temperature/top_p/top_k return a 400. Same tokenizer as Sonnet 5. 1M context window, 128K max output.";
+    readonly maxInputTokens: 1000000;
+    readonly maxOutputTokens: 128000;
+    readonly inputTokenCost: 2;
+    readonly cachedInputTokenCost: 0.2;
+    readonly cacheCreationInputTokenCost: 2.5;
+    readonly outputTokenCost: 10;
+    readonly reasoning: {
+        readonly thinkingStyle: "adaptive";
+        readonly levels: readonly ["low", "medium", "high", "xhigh", "max"];
+        readonly defaultLevel: "high";
+        readonly canDisable: false;
+        readonly outputsThinking: true;
+        readonly outputsSignatures: true;
+    };
+    readonly modalities: {
+        readonly input: readonly ["text", "image", "pdf"];
+        readonly output: readonly ["text"];
+    };
+    readonly knowledge: "2026-06";
+    readonly releaseDate: "2026-09-28";
+    readonly lastUpdated: "2026-09-28";
+    readonly family: "claude-sonnet";
+    readonly openWeights: false;
+    readonly temperatureSupported: false;
+    readonly provider: "anthropic";
 }, {
     readonly type: "text";
     readonly modelName: "claude-sonnet-5";
-    readonly description: "The best combination of speed and intelligence in the Sonnet tier, with near-Opus quality on coding and agentic work. Adaptive thinking on by default; supports the full low/medium/high/xhigh/max effort range. New tokenizer (~30% more tokens for the same text vs Sonnet 4.6). The $2/$10 per MTok launch rate is now the standard price — the increase to $3/$15 scheduled for 2026-09-01 was cancelled. 1M context window, 128K max output.";
+    readonly description: "Legacy Sonnet-tier model, superseded by claude-sonnet-5-5 at the same price. Near-Opus quality on coding and agentic work. Adaptive thinking on by default; supports the full low/medium/high/xhigh/max effort range. New tokenizer (~30% more tokens for the same text vs Sonnet 4.6). The $2/$10 per MTok launch rate is now the standard price — the increase to $3/$15 scheduled for 2026-09-01 was cancelled. 1M context window, 128K max output.";
     readonly maxInputTokens: 1000000;
     readonly maxOutputTokens: 128000;
     readonly inputTokenCost: 2;
@@ -1716,6 +1917,8 @@ export declare const textModels: readonly [{
     readonly outputTokensPerSecond: 52;
     readonly reasoning: {
         readonly thinkingStyle: "adaptive";
+        readonly levels: readonly ["low", "medium", "high", "max"];
+        readonly defaultLevel: "high";
         readonly canDisable: true;
         readonly outputsThinking: true;
         readonly outputsSignatures: true;
@@ -1797,6 +2000,7 @@ export declare const textModels: readonly [{
     readonly cachedInputTokenCost: 0.1;
     readonly cacheCreationInputTokenCost: 1.25;
     readonly outputTokenCost: 5;
+    readonly outputTokensPerSecond: 97;
     readonly reasoning: {
         readonly thinkingStyle: "budget";
         readonly canDisable: true;
@@ -1817,13 +2021,14 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "claude-opus-4-5";
-    readonly description: "Claude Opus 4.5 — earlier Opus generation. Prefer claude-opus-4-8.";
+    readonly description: "Claude Opus 4.5 — earlier Opus generation. Prefer claude-opus-5-5. Supports the effort parameter alongside budget_tokens thinking.";
     readonly maxInputTokens: 200000;
     readonly maxOutputTokens: 64000;
     readonly inputTokenCost: 5;
     readonly cachedInputTokenCost: 0.5;
     readonly cacheCreationInputTokenCost: 6.25;
     readonly outputTokenCost: 25;
+    readonly outputTokensPerSecond: 49;
     readonly reasoning: {
         readonly thinkingStyle: "budget";
         readonly canDisable: true;
@@ -1844,13 +2049,14 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "claude-sonnet-4-5";
-    readonly description: "Claude Sonnet 4.5 — earlier Sonnet generation. Prefer claude-sonnet-4-6.";
+    readonly description: "Claude Sonnet 4.5 — earlier Sonnet generation. Deprecated 2026-09-30; retires November 30, 2026. Migrate to claude-sonnet-5-5.";
     readonly maxInputTokens: 200000;
     readonly maxOutputTokens: 64000;
     readonly inputTokenCost: 3;
     readonly cachedInputTokenCost: 0.3;
     readonly cacheCreationInputTokenCost: 3.75;
     readonly outputTokenCost: 15;
+    readonly outputTokensPerSecond: 46;
     readonly reasoning: {
         readonly thinkingStyle: "budget";
         readonly canDisable: true;
@@ -1871,7 +2077,7 @@ export declare const textModels: readonly [{
 }, {
     readonly type: "text";
     readonly modelName: "gpt-5-pro";
-    readonly description: "GPT-5 Pro uses more compute for complex reasoning tasks. Available via the Responses API.";
+    readonly description: "GPT-5 Pro uses more compute for complex reasoning tasks. Available via the Responses API. Scheduled shutdown December 11, 2026 (announced 2026-06-11); migrate to gpt-5.6-sol with pro reasoning.";
     readonly maxInputTokens: 400000;
     readonly maxOutputTokens: 272000;
     readonly inputTokenCost: 15;
@@ -1913,17 +2119,46 @@ export declare const imageModels: readonly [{
     readonly type: "image";
     readonly modelName: "gpt-image-1";
     readonly provider: "openai";
+    readonly description: "Scheduled shutdown October 23, 2026; migrate to gpt-image-2.5-sunburst or gpt-image-2.5-flare.";
     readonly inputTokenCost: 5;
     readonly cachedInputTokenCost: 1.25;
     readonly inputImageTokenCost: 10;
     readonly outputImageTokenCost: 40;
     readonly costPerImage: 0.25;
+}, {
+    readonly type: "image";
+    readonly modelName: "gpt-image-2";
+    readonly provider: "openai";
+    readonly description: "Replacement for gpt-image-1-mini, gpt-image-1.5 and chatgpt-image-latest (those shut down 2026-12-01).";
+    readonly inputTokenCost: 5;
+    readonly cachedInputTokenCost: 1.25;
+    readonly inputImageTokenCost: 8;
+    readonly outputImageTokenCost: 30;
+}, {
+    readonly type: "image";
+    readonly modelName: "gpt-image-2.5-sunburst";
+    readonly provider: "openai";
+    readonly description: "GPT Image 2.5 Sunburst (released 2026-09-08). Quality low/medium/high/xhigh/max/auto. Cached image input $2/1M; text output is not billed.";
+    readonly inputTokenCost: 5;
+    readonly cachedInputTokenCost: 1.25;
+    readonly inputImageTokenCost: 8;
+    readonly outputImageTokenCost: 30;
+}, {
+    readonly type: "image";
+    readonly modelName: "gpt-image-2.5-flare";
+    readonly provider: "openai";
+    readonly description: "GPT Image 2.5 Flare (released 2026-09-08). Quality low/medium/high/xhigh/max/auto. Cached image input $2/1M; text output is not billed.";
+    readonly inputTokenCost: 5;
+    readonly cachedInputTokenCost: 1.25;
+    readonly inputImageTokenCost: 8;
+    readonly outputImageTokenCost: 30;
 }, {
     readonly type: "image";
     readonly modelName: "gemini-2.5-flash-image";
     readonly provider: "google";
-    readonly description: "aka nano-banana. Graduated from preview (was gemini-2.5-flash-image-preview).";
-    readonly costPerImage: 0.04;
+    readonly description: "aka nano-banana. RETIRED: shut down October 2, 2026. Use gemini-3.1-flash-image or gemini-3.1-flash-lite-image.";
+    readonly costPerImage: 0.039;
+    readonly disabled: true;
 }, {
     readonly type: "image";
     readonly modelName: "nano-banana-pro-preview";
@@ -1934,8 +2169,15 @@ export declare const imageModels: readonly [{
     readonly type: "image";
     readonly modelName: "gemini-3-pro-image-preview";
     readonly provider: "google";
-    readonly description: "High-fidelity image generation with reasoning-enhanced composition. Supports legible text rendering, complex multi-turn editing, and character consistency using up to 14 reference inputs.";
+    readonly description: "DEPRECATED: Preview version, shut down June 25, 2026. Use gemini-3-pro-image instead.";
     readonly costPerImage: 0.05;
+    readonly disabled: true;
+}, {
+    readonly type: "image";
+    readonly modelName: "gemini-3-pro-image";
+    readonly provider: "google";
+    readonly description: "aka Nano Banana Pro (GA 2026-05-28). High-fidelity image generation with reasoning-enhanced composition. Supports legible text rendering, complex multi-turn editing, and character consistency using up to 14 reference inputs. $0.134/image at 1K or 2K, $0.24 at 4K; text input $2/1M, text/thinking output $12/1M.";
+    readonly costPerImage: 0.134;
 }, {
     readonly type: "image";
     readonly modelName: "gemini-3.1-flash-image-preview";
@@ -1953,7 +2195,7 @@ export declare const imageModels: readonly [{
     readonly type: "image";
     readonly modelName: "gemini-3.1-flash-lite-image";
     readonly provider: "google";
-    readonly description: "aka Nano Banana 2 Lite (GA 2026-06-30). Fastest, most cost-effective Gemini image model (~4s generation). ~$0.034/image at 1K. Recommended replacement for gemini-2.5-flash-image.";
+    readonly description: "aka Nano Banana 2 Lite (GA 2026-06-30). Fastest, most cost-effective Gemini image model (~4s generation). ~$0.034/image at 1K.";
     readonly costPerImage: 0.034;
 }];
 export declare const embeddingsModels: EmbeddingsModel[];