@botpress/cognitive 0.4.7 → 0.5.0

package/dist/index.mjs

@@ -646,10 +646,79 @@ var models = {
     tags: ["recommended", "reasoning", "general-purpose", "vision", "coding", "agents"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: true
     },
     aliases: ["gpt-5.4"]
   },
+  "openai:gpt-5.4-mini-2026-03-17": {
+    id: "openai:gpt-5.4-mini-2026-03-17",
+    name: "GPT-5.4 Mini",
+    description: "GPT-5.4 Mini brings the strengths of GPT-5.4 to a faster, more efficient model designed for high-volume workloads. It is optimized for speed and cost while retaining strong reasoning and vision capabilities.",
+    input: {
+      maxTokens: 4e5,
+      costPer1MTokens: 0.75
+    },
+    output: {
+      maxTokens: 128e3,
+      costPer1MTokens: 4.5
+    },
+    tags: ["recommended", "reasoning", "general-purpose", "vision"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    },
+    aliases: ["gpt-5.4-mini"]
+  },
+  "openai:gpt-5.4-nano-2026-03-17": {
+    id: "openai:gpt-5.4-nano-2026-03-17",
+    name: "GPT-5.4 Nano",
+    description: "GPT-5.4 Nano is the smallest and cheapest GPT-5.4 variant, designed for tasks where speed and cost matter most like classification, data extraction, ranking, and coding sub-agents.",
+    input: {
+      maxTokens: 4e5,
+      costPer1MTokens: 0.2
+    },
+    output: {
+      maxTokens: 128e3,
+      costPer1MTokens: 1.25
+    },
+    tags: ["low-cost", "reasoning", "general-purpose", "vision"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    },
+    aliases: ["gpt-5.4-nano"]
+  },
+  "openai:gpt-5.3-chat": {
+    id: "openai:gpt-5.3-chat",
+    name: "GPT-5.3 Chat",
+    description: "GPT-5.3 Chat is the GPT-5.3 Instant model used in ChatGPT, exposed via the API. Rolling alias that points to the latest snapshot.",
+    input: {
+      maxTokens: 128e3,
+      costPer1MTokens: 1.75
+    },
+    output: {
+      maxTokens: 16384,
+      costPer1MTokens: 14
+    },
+    tags: ["reasoning", "general-purpose", "vision"],
+    lifecycle: "preview",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    },
+    aliases: ["gpt-5.3-chat-latest"]
+  },
   "openai:gpt-5.2-2025-12-11": {
     id: "openai:gpt-5.2-2025-12-11",
     name: "GPT-5.2",
@@ -665,7 +734,10 @@ var models = {
     tags: ["recommended", "reasoning", "general-purpose", "vision"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: true
     }
   },
   "openai:gpt-5.1-2025-11-13": {
@@ -683,7 +755,10 @@ var models = {
     tags: ["recommended", "reasoning", "general-purpose", "vision"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: true
     }
   },
   "openai:gpt-5-2025-08-07": {
@@ -701,7 +776,10 @@ var models = {
     tags: ["reasoning", "general-purpose", "vision"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["gpt-5"]
   },
@@ -720,7 +798,10 @@ var models = {
     tags: ["recommended", "reasoning", "general-purpose", "vision"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["gpt-5-mini"]
   },
@@ -739,7 +820,10 @@ var models = {
     tags: ["low-cost", "reasoning", "general-purpose", "vision"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["gpt-5-nano"]
   },
@@ -758,7 +842,10 @@ var models = {
     tags: ["reasoning", "vision", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["o4-mini"]
   },
@@ -777,7 +864,10 @@ var models = {
     tags: ["reasoning", "vision", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["o3"]
   },
@@ -796,7 +886,10 @@ var models = {
     tags: ["recommended", "vision", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["gpt-4.1"]
   },
@@ -815,7 +908,10 @@ var models = {
     tags: ["recommended", "vision", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["gpt-4.1-mini"]
   },
@@ -834,7 +930,10 @@ var models = {
     tags: ["low-cost", "vision", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["gpt-4.1-nano"]
   },
@@ -853,7 +952,10 @@ var models = {
     tags: ["reasoning", "general-purpose", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["o3-mini"]
   },
@@ -872,7 +974,10 @@ var models = {
     tags: ["reasoning", "vision", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "openai:o1-mini-2024-09-12": {
@@ -888,9 +993,12 @@ var models = {
       costPer1MTokens: 4.4
     },
     tags: ["reasoning", "vision", "general-purpose"],
-    lifecycle: "production",
+    lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["o1-mini"]
   },
@@ -909,7 +1017,10 @@ var models = {
     tags: ["recommended", "vision", "low-cost", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["gpt-4o-mini"]
   },
@@ -928,7 +1039,10 @@ var models = {
     tags: ["recommended", "vision", "general-purpose", "coding", "agents"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["gpt-4o"]
   },
@@ -947,7 +1061,10 @@ var models = {
     tags: ["deprecated", "vision", "general-purpose", "coding", "agents"],
     lifecycle: "deprecated",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "openai:gpt-4o-2024-05-13": {
@@ -965,7 +1082,10 @@ var models = {
     tags: ["deprecated", "vision", "general-purpose", "coding", "agents"],
     lifecycle: "deprecated",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "openai:gpt-4-turbo-2024-04-09": {
@@ -983,7 +1103,10 @@ var models = {
     tags: ["deprecated", "general-purpose", "coding", "agents"],
     lifecycle: "deprecated",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "openai:gpt-3.5-turbo-0125": {
@@ -1001,7 +1124,32 @@ var models = {
     tags: ["deprecated", "general-purpose", "low-cost"],
     lifecycle: "deprecated",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    }
+  },
+  "openai:whisper-1": {
+    id: "openai:whisper-1",
+    name: "Whisper V2",
+    description: "OpenAI Whisper V2 \u2014 general-purpose speech recognition model supporting 99 languages.",
+    input: {
+      maxTokens: 0,
+      costPer1MTokens: 0,
+      costPerMinute: 6e-3
+    },
+    output: {
+      maxTokens: 0,
+      costPer1MTokens: 0
+    },
+    tags: ["general-purpose", "speech-to-text"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: true,
+      supportsSearch: false
     }
   },
   "anthropic:claude-opus-4-6": {
@@ -1009,7 +1157,7 @@ var models = {
     name: "Claude Opus 4.6",
     description: "Claude Opus 4.6 is the most intelligent Claude model, built for complex agents and coding workflows. It excels at long-running professional tasks, large codebases, complex refactors, and multi-step debugging with a 128K max output.",
     input: {
-      maxTokens: 2e5,
+      maxTokens: 1e6,
       costPer1MTokens: 5
     },
     output: {
@@ -1019,7 +1167,10 @@ var models = {
     tags: ["recommended", "reasoning", "agents", "vision", "general-purpose", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: true
     }
   },
   "anthropic:claude-sonnet-4-6": {
@@ -1027,7 +1178,7 @@ var models = {
     name: "Claude Sonnet 4.6",
     description: "Claude Sonnet 4.6 offers the best combination of speed and intelligence in the Claude family. It features adaptive thinking for dynamic reasoning allocation, delivering fast responses for simple queries and deeper analysis for complex tasks.",
     input: {
-      maxTokens: 2e5,
+      maxTokens: 1e6,
       costPer1MTokens: 3
     },
     output: {
@@ -1037,9 +1188,34 @@ var models = {
     tags: ["recommended", "reasoning", "agents", "vision", "general-purpose", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: true
     }
   },
+  "anthropic:claude-opus-4-5-20251101": {
+    id: "anthropic:claude-opus-4-5-20251101",
+    name: "Claude Opus 4.5",
+    description: "Claude Opus 4.5 is a highly capable model with strong reasoning, coding, and agentic performance. It offers the same pricing tier as Opus 4.6 with a 200K context window.",
+    input: {
+      maxTokens: 2e5,
+      costPer1MTokens: 5
+    },
+    output: {
+      maxTokens: 64e3,
+      costPer1MTokens: 25
+    },
+    tags: ["reasoning", "agents", "vision", "general-purpose", "coding"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: true
+    },
+    aliases: ["claude-opus-4-5"]
+  },
   "anthropic:claude-sonnet-4-5-20250929": {
     id: "anthropic:claude-sonnet-4-5-20250929",
     name: "Claude Sonnet 4.5",
@@ -1055,7 +1231,10 @@ var models = {
     tags: ["recommended", "reasoning", "agents", "vision", "general-purpose", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: true
     },
     aliases: ["claude-sonnet-4-5"]
   },
@@ -1074,7 +1253,10 @@ var models = {
     tags: ["recommended", "reasoning", "agents", "vision", "general-purpose", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["claude-sonnet-4"]
   },
@@ -1093,7 +1275,10 @@ var models = {
     tags: ["deprecated", "vision", "reasoning", "general-purpose", "agents", "coding"],
     lifecycle: "deprecated",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["claude-sonnet-4-reasoning"]
   },
@@ -1112,7 +1297,10 @@ var models = {
     tags: ["recommended", "agents", "vision", "general-purpose", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["claude-haiku-4-5"]
   },
@@ -1133,9 +1321,12 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["recommended", "reasoning", "agents", "vision", "general-purpose", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
-    aliases: ["claude-haiku-4-5-reasoning"]
+    aliases: ["claude-haiku-4-5-reasoning", "claude-haiku-4-5-20251001"]
   },
   "anthropic:claude-3-7-sonnet-20250219": {
     id: "anthropic:claude-3-7-sonnet-20250219",
@@ -1152,7 +1343,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["recommended", "reasoning", "agents", "vision", "general-purpose", "coding"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "anthropic:claude-3-7-sonnet-reasoning-20250219": {
@@ -1170,7 +1364,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["deprecated", "vision", "reasoning", "general-purpose", "agents", "coding"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "anthropic:claude-3-5-haiku-20241022": {
@@ -1188,7 +1385,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["general-purpose", "low-cost"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "anthropic:claude-3-5-sonnet-20241022": {
@@ -1206,7 +1406,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["vision", "general-purpose", "agents", "coding", "storytelling"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "anthropic:claude-3-5-sonnet-20240620": {
@@ -1224,7 +1427,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["vision", "general-purpose", "agents", "coding", "storytelling"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "anthropic:claude-3-haiku-20240307": {
@@ -1242,13 +1448,16 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["deprecated", "low-cost", "general-purpose"],
     lifecycle: "deprecated",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
-  "google-ai:gemini-3-pro": {
-    id: "google-ai:gemini-3-pro",
-    name: "Gemini 3 Pro",
-    description: "One of the best models for multimodal understanding, and Google's most powerful agentic and vibe-coding model yet, delivering richer visuals and deeper interactivity, built on a foundation of state-of-the-art reasoning.",
+  "google-ai:gemini-3.1-pro": {
+    id: "google-ai:gemini-3.1-pro",
+    name: "Gemini 3.1 Pro",
+    description: "Google's most powerful agentic and coding model, delivering state-of-the-art reasoning with rich multimodal understanding. Successor to Gemini 3 Pro (shut down March 9, 2026).",
     input: {
       maxTokens: 1048576,
       costPer1MTokens: 2
@@ -1257,10 +1466,13 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       maxTokens: 65536,
       costPer1MTokens: 12
     },
-    tags: ["reasoning", "agents", "general-purpose", "vision"],
+    tags: ["reasoning", "agents", "general-purpose", "vision", "coding"],
     lifecycle: "preview",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: true,
+      supportsTranscription: false,
+      supportsSearch: true
     },
     aliases: ["gemini-3.1-pro-preview"]
   },
@@ -1279,10 +1491,56 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["reasoning", "agents", "general-purpose", "vision"],
     lifecycle: "preview",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: true,
+      supportsTranscription: false,
+      supportsSearch: true
     },
     aliases: ["gemini-3-flash-preview"]
   },
+  "google-ai:gemini-3.1-flash-lite": {
+    id: "google-ai:gemini-3.1-flash-lite",
+    name: "Gemini 3.1 Flash-Lite",
+    description: "Google's most cost-effective AI model for high-volume, low-latency tasks. Offers strong performance at a fraction of the cost of larger models.",
+    input: {
+      maxTokens: 1048576,
+      costPer1MTokens: 0.25
+    },
+    output: {
+      maxTokens: 65536,
+      costPer1MTokens: 1.5
+    },
+    tags: ["low-cost", "general-purpose", "vision"],
+    lifecycle: "preview",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: true,
+      supportsTranscription: false,
+      supportsSearch: true
+    },
+    aliases: ["gemini-3.1-flash-lite-preview"]
+  },
+  "google-ai:gemini-2.5-pro": {
+    id: "google-ai:gemini-2.5-pro",
+    name: "Gemini 2.5 Pro",
+    description: `Google's most advanced stable AI model designed for complex reasoning, coding, mathematics, and scientific tasks. Features "thinking" capabilities for superior human-preference alignment and problem-solving.`,
+    input: {
+      maxTokens: 2e5,
+      costPer1MTokens: 1.25
+    },
+    output: {
+      maxTokens: 65536,
+      costPer1MTokens: 10
+    },
+    tags: ["recommended", "reasoning", "agents", "general-purpose", "vision", "coding"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: true,
+      supportsTranscription: false,
+      supportsSearch: true
+    }
+  },
   "google-ai:gemini-2.5-flash": {
     id: "google-ai:gemini-2.5-flash",
     name: "Gemini 2.5 Flash",
@@ -1298,25 +1556,31 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["recommended", "reasoning", "agents", "general-purpose", "vision"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: true,
+      supportsTranscription: false,
+      supportsSearch: true
     }
   },
-  "google-ai:gemini-2.5-pro": {
-    id: "google-ai:gemini-2.5-pro",
-    name: "Gemini 2.5 Pro",
-    description: `Google's most advanced AI model designed for complex reasoning, coding, mathematics, and scientific tasks. Features "thinking" capabilities for superior human-preference alignment and problem-solving.`,
+  "google-ai:gemini-2.5-flash-lite": {
+    id: "google-ai:gemini-2.5-flash-lite",
+    name: "Gemini 2.5 Flash-Lite",
+    description: "Lightweight, cost-efficient Gemini model optimized for high-volume, low-latency tasks. Successor to Gemini 2.0 Flash with improved capabilities.",
     input: {
-      maxTokens: 2e5,
-      costPer1MTokens: 1.25
+      maxTokens: 1048576,
+      costPer1MTokens: 0.1
     },
     output: {
       maxTokens: 65536,
-      costPer1MTokens: 10
+      costPer1MTokens: 0.4
     },
-    tags: ["recommended", "reasoning", "agents", "general-purpose", "vision", "coding"],
+    tags: ["recommended", "low-cost", "general-purpose", "vision"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: true,
+      supportsTranscription: false,
+      supportsSearch: true
     }
   },
   "google-ai:gemini-2.0-flash": {
@@ -1332,12 +1596,37 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       costPer1MTokens: 0.4
     },
     tags: ["low-cost", "general-purpose", "vision"],
-    lifecycle: "production",
+    lifecycle: "deprecated",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: true,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["models/gemini-2.0-flash"]
   },
+  "google-ai:gemini-3-pro": {
+    id: "google-ai:gemini-3-pro",
+    name: "Gemini 3 Pro (Shut Down)",
+    description: "Gemini 3 Pro Preview was shut down on March 9, 2026. Use Gemini 3.1 Pro instead.",
+    input: {
+      maxTokens: 1048576,
+      costPer1MTokens: 2
+    },
+    output: {
+      maxTokens: 65536,
+      costPer1MTokens: 12
+    },
+    tags: ["reasoning", "agents", "general-purpose", "vision"],
+    lifecycle: "discontinued",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: true,
+      supportsTranscription: false,
+      supportsSearch: true
+    },
+    aliases: ["gemini-3-pro-preview"]
+  },
   "cerebras:gpt-oss-120b": {
     id: "cerebras:gpt-oss-120b",
     name: "GPT-OSS 120B (Preview)",
@@ -1353,7 +1642,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["preview", "general-purpose", "reasoning"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "cerebras:qwen-3-32b": {
@@ -1369,9 +1661,12 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       costPer1MTokens: 0.8
     },
     tags: ["general-purpose", "reasoning"],
-    lifecycle: "production",
+    lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "cerebras:llama-4-scout-17b-16e-instruct": {
@@ -1387,9 +1682,12 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       costPer1MTokens: 0.85
     },
     tags: ["general-purpose", "vision"],
-    lifecycle: "production",
+    lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "cerebras:llama3.1-8b": {
@@ -1407,7 +1705,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["low-cost", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "cerebras:llama3.3-70b": {
@@ -1423,9 +1724,12 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       costPer1MTokens: 1.2
     },
     tags: ["general-purpose"],
-    lifecycle: "production",
+    lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:gpt-oss-20b": {
@@ -1434,16 +1738,19 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     description: "gpt-oss-20b is a compact, open-weight language model optimized for low-latency. It shares the same training foundation and capabilities as the GPT-OSS 120B model, with faster responses and lower cost.",
     input: {
       maxTokens: 131e3,
-      costPer1MTokens: 0.1
+      costPer1MTokens: 0.075
     },
     output: {
       maxTokens: 32e3,
-      costPer1MTokens: 0.5
+      costPer1MTokens: 0.3
     },
     tags: ["preview", "general-purpose", "reasoning", "low-cost"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["openai/gpt-oss-20b"]
   },
@@ -1462,7 +1769,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["preview", "general-purpose", "reasoning"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["openai/gpt-oss-120b"]
   },
@@ -1481,7 +1791,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["general-purpose", "reasoning", "preview"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:llama-3.3-70b-versatile": {
@@ -1499,7 +1812,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["recommended", "general-purpose", "coding"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:llama-3.2-1b-preview": {
@@ -1517,7 +1833,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["low-cost", "deprecated"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:llama-3.2-3b-preview": {
@@ -1535,7 +1854,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["low-cost", "general-purpose", "deprecated"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:llama-3.2-11b-vision-preview": {
@@ -1553,7 +1875,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["low-cost", "vision", "general-purpose", "deprecated"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:llama-3.2-90b-vision-preview": {
@@ -1571,7 +1896,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["vision", "general-purpose", "deprecated"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:llama-3.1-8b-instant": {
@@ -1589,7 +1917,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["low-cost", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:llama3-8b-8192": {
@@ -1607,7 +1938,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["low-cost", "general-purpose", "deprecated"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:llama3-70b-8192": {
@@ -1625,7 +1959,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["general-purpose", "deprecated"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "groq:gemma2-9b-it": {
@@ -1643,7 +1980,160 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["low-cost", "general-purpose"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    }
+  },
+  "groq:whisper-large-v3": {
+    id: "groq:whisper-large-v3",
+    name: "Whisper V3",
+    description: "Whisper Large V3 on Groq \u2014 fast, accurate multilingual speech recognition.",
+    input: {
+      maxTokens: 0,
+      costPer1MTokens: 0,
+      costPerMinute: 185e-5
+    },
+    output: {
+      maxTokens: 0,
+      costPer1MTokens: 0
+    },
+    tags: ["general-purpose", "speech-to-text"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: true,
+      supportsSearch: false
+    }
+  },
+  "groq:whisper-large-v3-turbo": {
+    id: "groq:whisper-large-v3-turbo",
+    name: "Whisper V3 Turbo",
+    description: "Whisper Large V3 Turbo on Groq \u2014 optimized for speed with near-identical accuracy to V3.",
+    input: {
+      maxTokens: 0,
+      costPer1MTokens: 0,
+      costPerMinute: 667e-6
+    },
+    output: {
+      maxTokens: 0,
+      costPer1MTokens: 0
+    },
+    tags: ["low-cost", "general-purpose", "speech-to-text"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: true,
+      supportsSearch: false
+    }
+  },
+  "groq:distil-whisper-large-v3-en": {
+    id: "groq:distil-whisper-large-v3-en",
+    name: "Distil Whisper V3 (English)",
+    description: "Distilled Whisper Large V3 on Groq \u2014 decommissioned, replaced by whisper-large-v3-turbo.",
+    input: {
+      maxTokens: 0,
+      costPer1MTokens: 0,
+      costPerMinute: 333e-6
+    },
+    output: {
+      maxTokens: 0,
+      costPer1MTokens: 0
+    },
+    tags: ["low-cost", "speech-to-text"],
+    lifecycle: "discontinued",
+    capabilities: {
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: true,
+      supportsSearch: false
+    }
+  },
+  "xai:grok-4-1-fast-reasoning": {
+    id: "xai:grok-4-1-fast-reasoning",
+    name: "Grok 4.1 Fast (Reasoning)",
+    description: "Latest fast Grok model with reasoning capabilities and a massive 2M context window. Extremely cost-effective for a frontier-class model.",
+    input: {
+      maxTokens: 2e6,
+      costPer1MTokens: 0.2
+    },
+    output: {
+      maxTokens: 128e3,
+      costPer1MTokens: 0.5
+    },
+    tags: ["recommended", "reasoning", "general-purpose", "vision", "low-cost"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    }
+  },
+  "xai:grok-4-1-fast-non-reasoning": {
+    id: "xai:grok-4-1-fast-non-reasoning",
+    name: "Grok 4.1 Fast (Non-Reasoning)",
+    description: "Latest fast Grok model for non-reasoning tasks with a massive 2M context window. Extremely cost-effective for a frontier-class model.",
+    input: {
+      maxTokens: 2e6,
+      costPer1MTokens: 0.2
+    },
+    output: {
+      maxTokens: 128e3,
+      costPer1MTokens: 0.5
+    },
+    tags: ["recommended", "general-purpose", "vision", "low-cost"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    }
+  },
+  "xai:grok-4.20-0309-reasoning": {
+    id: "xai:grok-4.20-0309-reasoning",
+    name: "Grok 4.20 (Reasoning)",
+    description: "xAI flagship model with deep reasoning capabilities and 2M context window.",
+    input: {
+      maxTokens: 2e6,
+      costPer1MTokens: 2
+    },
+    output: {
+      maxTokens: 128e3,
+      costPer1MTokens: 6
+    },
+    tags: ["reasoning", "general-purpose", "vision"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    }
+  },
+  "xai:grok-4.20-0309-non-reasoning": {
+    id: "xai:grok-4.20-0309-non-reasoning",
+    name: "Grok 4.20 (Non-Reasoning)",
+    description: "xAI flagship model for non-reasoning tasks with 2M context window.",
+    input: {
+      maxTokens: 2e6,
+      costPer1MTokens: 2
+    },
+    output: {
+      maxTokens: 128e3,
+      costPer1MTokens: 6
+    },
+    tags: ["general-purpose", "vision"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "xai:grok-code-fast-1": {
@@ -1661,7 +2151,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["coding", "general-purpose", "low-cost"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "xai:grok-4-fast-reasoning": {
@@ -1676,10 +2169,13 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       maxTokens: 128e3,
       costPer1MTokens: 0.5
     },
-    tags: ["reasoning", "recommended", "general-purpose"],
+    tags: ["reasoning", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "xai:grok-4-fast-non-reasoning": {
@@ -1694,10 +2190,13 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       maxTokens: 128e3,
       costPer1MTokens: 0.5
     },
-    tags: ["low-cost", "recommended", "general-purpose"],
+    tags: ["low-cost", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "xai:grok-4-0709": {
@@ -1715,7 +2214,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["reasoning", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "xai:grok-3-mini": {
@@ -1733,7 +2235,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["low-cost", "general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "xai:grok-3": {
@@ -1751,7 +2256,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
   "openrouter:gpt-oss-120b": {
@@ -1769,9 +2277,34 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["preview", "general-purpose", "reasoning"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     }
   },
+  "fireworks-ai:qwen3-8b": {
+    id: "fireworks-ai:qwen3-8b",
+    name: "Qwen3 8B",
+    description: "Qwen3 8B is a newer-generation small model with better architecture than Llama 3.1 8B. Same price tier with improved quality across reasoning and coding tasks.",
+    input: {
+      maxTokens: 40960,
+      costPer1MTokens: 0.2
+    },
+    output: {
+      maxTokens: 16384,
+      costPer1MTokens: 0.2
+    },
+    tags: ["low-cost", "general-purpose", "reasoning"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    },
+    aliases: ["accounts/fireworks/models/qwen3-8b"]
+  },
   "fireworks-ai:gpt-oss-20b": {
     id: "fireworks-ai:gpt-oss-20b",
     name: "GPT-OSS 20B",
@@ -1787,7 +2320,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["general-purpose", "reasoning", "low-cost"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/gpt-oss-20b"]
   },
@@ -1806,10 +2342,57 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["general-purpose", "reasoning"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/gpt-oss-120b"]
   },
+  "fireworks-ai:deepseek-v3p2": {
+    id: "fireworks-ai:deepseek-v3p2",
+    name: "DeepSeek V3.2",
+    description: "DeepSeek V3.2 is a 675B-parameter mixture-of-experts model with superior reasoning and agent performance. It delivers high computational efficiency with strong results across coding, math, and general-purpose tasks.",
+    input: {
+      maxTokens: 163840,
+      costPer1MTokens: 0.56
+    },
+    output: {
+      maxTokens: 16e4,
+      costPer1MTokens: 1.68
+    },
+    tags: ["recommended", "reasoning", "general-purpose", "coding"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    },
+    aliases: ["accounts/fireworks/models/deepseek-v3p2"]
+  },
+  "fireworks-ai:deepseek-v3p1": {
+    id: "fireworks-ai:deepseek-v3p1",
+    name: "DeepSeek V3.1",
+    description: "DeepSeek V3.1 is a 685B-parameter hybrid LLM with mixture-of-experts architecture (37B activated per token). Features thinking and non-thinking chat modes for complex agentic behaviors and reasoning tasks.",
+    input: {
+      maxTokens: 163840,
+      costPer1MTokens: 0.56
+    },
+    output: {
+      maxTokens: 163840,
+      costPer1MTokens: 1.68
+    },
+    tags: ["reasoning", "general-purpose", "coding"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
+    },
+    aliases: ["accounts/fireworks/models/deepseek-v3p1"]
+  },
   "fireworks-ai:deepseek-r1-0528": {
     id: "fireworks-ai:deepseek-r1-0528",
     name: "DeepSeek R1 0528",
@@ -1823,9 +2406,12 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       costPer1MTokens: 8
     },
     tags: ["recommended", "reasoning", "general-purpose", "coding"],
-    lifecycle: "production",
+    lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/deepseek-r1-0528"]
   },
@@ -1842,9 +2428,12 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       costPer1MTokens: 0.9
     },
     tags: ["recommended", "general-purpose"],
-    lifecycle: "production",
+    lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/deepseek-v3-0324"]
   },
@@ -1861,9 +2450,12 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       costPer1MTokens: 0.88
     },
     tags: ["general-purpose", "vision"],
-    lifecycle: "production",
+    lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/llama4-maverick-instruct-basic"]
   },
@@ -1880,9 +2472,12 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       costPer1MTokens: 0.6
     },
     tags: ["general-purpose", "vision"],
-    lifecycle: "production",
+    lifecycle: "discontinued",
     capabilities: {
-      supportsImages: true
+      supportsImages: true,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/llama4-scout-instruct-basic"]
   },
@@ -1901,7 +2496,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["general-purpose"],
     lifecycle: "production",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/llama-v3p3-70b-instruct"]
   },
@@ -1920,7 +2518,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["reasoning", "general-purpose", "coding"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/deepseek-r1"]
   },
@@ -1939,7 +2540,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["reasoning", "general-purpose", "coding"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/deepseek-r1-basic"]
   },
@@ -1958,7 +2562,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["deprecated", "general-purpose"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/deepseek-v3"]
   },
@@ -1977,7 +2584,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["deprecated", "general-purpose"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/llama-v3p1-405b-instruct"]
   },
@@ -1996,7 +2606,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["deprecated", "general-purpose"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/llama-v3p1-70b-instruct"]
   },
@@ -2013,9 +2626,12 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
       costPer1MTokens: 0.2
     },
     tags: ["low-cost", "general-purpose"],
-    lifecycle: "production",
+    lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/llama-v3p1-8b-instruct"]
   },
@@ -2034,7 +2650,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["deprecated", "general-purpose"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/mixtral-8x22b-instruct"]
   },
@@ -2053,7 +2672,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["low-cost", "general-purpose"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/mixtral-8x7b-instruct"]
   },
@@ -2072,7 +2694,10 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["roleplay", "storytelling", "low-cost"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/mythomax-l2-13b"]
   },
@@ -2091,9 +2716,34 @@ Claude Haiku 4.5 is Anthropic's fastest and most efficient model, delivering nea
     tags: ["deprecated", "low-cost", "general-purpose"],
     lifecycle: "discontinued",
     capabilities: {
-      supportsImages: false
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: false,
+      supportsSearch: false
     },
     aliases: ["accounts/fireworks/models/gemma2-9b-it"]
+  },
+  "fireworks-ai:whisper-v3": {
+    id: "fireworks-ai:whisper-v3",
+    name: "Whisper V3",
+    description: "Whisper V3 on Fireworks AI \u2014 multilingual speech recognition with high accuracy.",
+    input: {
+      maxTokens: 0,
+      costPer1MTokens: 0,
+      costPerMinute: 15e-4
+    },
+    output: {
+      maxTokens: 0,
+      costPer1MTokens: 0
+    },
+    tags: ["general-purpose", "speech-to-text"],
+    lifecycle: "production",
+    capabilities: {
+      supportsImages: false,
+      supportsAudio: false,
+      supportsTranscription: true,
+      supportsSearch: false
+    }
   }
 };
 var defaultModel = {
@@ -2157,7 +2807,7 @@ var CognitiveBeta = class _CognitiveBeta {
   }
   async generateText(input, options = {}) {
     const signal = options.signal ?? AbortSignal.timeout(this._timeout);
-    const req = { input };
+    const req = { type: "generateText", input };
     this._events.emit("request", req);
     try {
       const { data } = await this._withServerRetry(
@@ -2181,9 +2831,32 @@ var CognitiveBeta = class _CognitiveBeta {
     );
     return data.models;
   }
+  async transcribeAudio(input, options = {}) {
+    const signal = options.signal ?? AbortSignal.timeout(this._timeout);
+    const req = { type: "transcribeAudio", input };
+    this._events.emit("request", req);
+    try {
+      const { data } = await this._withServerRetry(
+        () => this._axiosClient.post("/v2/cognitive/transcribe-audio", input, {
+          signal,
+          timeout: options.timeout ?? this._timeout
+        }),
+        options,
+        req
+      );
+      if (data.error) {
+        throw new Error(`Transcription error: ${data.error}`);
+      }
+      this._events.emit("response", req, data);
+      return data;
+    } catch (error) {
+      this._events.emit("error", req, error);
+      throw error;
+    }
+  }
   async *generateTextStream(request, options = {}) {
     const signal = options.signal ?? AbortSignal.timeout(this._timeout);
-    const req = { input: request };
+    const req = { type: "generateText", input: request };
     const chunks = [];
     let lastChunk;
     this._events.emit("request", req);