diff --git a/packages/cognitive/src/features/providers/anthropic/anthropic.config.ts b/packages/cognitive/src/features/providers/anthropic/anthropic.config.ts
index 70853f9..4f41076 100644
--- a/packages/cognitive/src/features/providers/anthropic/anthropic.config.ts
+++ b/packages/cognitive/src/features/providers/anthropic/anthropic.config.ts
@@ -4,8 +4,128 @@ export const ANTHROPIC_CONFIG: ProviderConfig = {
   id: 'anthropic',
   name: 'Anthropic',
   description: 'Claude models focused on safety and helpfulness',
-  defaultModel: 'claude-sonnet-4-5-20250929',
+  defaultModel: 'claude-sonnet-4-6',
   models: [
+    {
+      id: 'claude-opus-4-6',
+      displayName: 'Claude Opus 4.6',
+      aliases: ['claude-opus-4-6'],
+      description:
+        "Claude Opus 4.6 is Anthropic's most capable model, featuring adaptive reasoning and exceptional performance across complex tasks. It delivers state-of-the-art results on coding, analysis, and agentic workflows, with a 1M token context window.",
+      lifecycle: 'production',
+      health: 'healthy',
+      releaseDate: '2026-01-20',
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: false,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 1_000_000,
+        maxOutputTokens: 128_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 5,
+        outputCostPer1mTokens: 25,
+      },
+      tags: ['recommended', 'reasoning', 'agents', 'vision', 'general-purpose', 'coding'],
+    },
+    {
+      id: 'claude-sonnet-4-6',
+      displayName: 'Claude Sonnet 4.6',
+      aliases: ['claude-sonnet-4-6'],
+      description:
+        "Claude Sonnet 4.6 is Anthropic's latest flagship Sonnet model, offering top-tier performance for agentic and coding workflows with a 1M token context window. It delivers state-of-the-art results on coding benchmarks and reasoning tasks.",
+      lifecycle: 'production',
+      health: 'healthy',
+      releaseDate: '2026-01-20',
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: false,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 1_000_000,
+        maxOutputTokens: 64_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 3,
+        outputCostPer1mTokens: 15,
+      },
+      tags: ['recommended', 'reasoning', 'agents', 'vision', 'general-purpose', 'coding'],
+    },
+    {
+      id: 'claude-opus-4-5-20251101',
+      displayName: 'Claude Opus 4.5',
+      aliases: ['claude-opus-4-5'],
+      description:
+        "Claude Opus 4.5 is Anthropic's high-performance Opus model, excelling in complex reasoning, coding, and long-context tasks.",
+      lifecycle: 'production',
+      health: 'healthy',
+      releaseDate: '2025-11-01',
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: false,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 200_000,
+        maxOutputTokens: 64_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 5,
+        outputCostPer1mTokens: 25,
+      },
+      tags: ['reasoning', 'agents', 'vision', 'general-purpose', 'coding'],
+    },
+    {
+      id: 'claude-opus-4-1-20250805',
+      displayName: 'Claude Opus 4.1',
+      aliases: ['claude-opus-4-1'],
+      description:
+        "Claude Opus 4.1 is a powerful Opus-tier model optimized for complex tasks requiring extended reasoning and high accuracy.",
+      lifecycle: 'production',
+      health: 'healthy',
+      releaseDate: '2025-08-05',
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: false,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 200_000,
+        maxOutputTokens: 32_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 15,
+        outputCostPer1mTokens: 75,
+      },
+      tags: ['reasoning', 'agents', 'vision', 'general-purpose', 'coding'],
+    },
     {
       id: 'claude-sonnet-4-5-20250929',
       displayName: 'Claude Sonnet 4.5',
@@ -36,6 +156,36 @@ export const ANTHROPIC_CONFIG: ProviderConfig = {
       },
       tags: ['recommended', 'reasoning', 'agents', 'vision', 'general-purpose', 'coding'],
     },
+    {
+      id: 'claude-opus-4-20250514',
+      displayName: 'Claude Opus 4',
+      aliases: ['claude-opus-4-0'],
+      description:
+        "Claude Opus 4 is Anthropic's frontier Opus model from May 2025, designed for complex reasoning, advanced coding, and long-horizon agentic tasks.",
+      lifecycle: 'production',
+      health: 'healthy',
+      releaseDate: '2025-05-14',
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: false,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 200_000,
+        maxOutputTokens: 32_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 15,
+        outputCostPer1mTokens: 75,
+      },
+      tags: ['reasoning', 'agents', 'vision', 'general-purpose', 'coding'],
+    },
     {
       id: 'claude-sonnet-4-20250514',
       displayName: 'Claude Sonnet 4',
@@ -320,9 +470,11 @@ export const ANTHROPIC_CONFIG: ProviderConfig = {
       displayName: 'Claude 3 Haiku',
       description:
         "Claude 3 Haiku is Anthropic's fastest and most compact model for near-instant responsiveness. Quick and accurate targeted performance.",
-      lifecycle: 'production',
+      lifecycle: 'deprecated',
       health: 'healthy',
       releaseDate: '2024-03-07',
+      deprecationDate: '2026-04-19',
+      replacementModels: ['claude-haiku-4-5-20251001'],
       capabilities: {
         supportsText: true,
         supportsImages: true,
diff --git a/packages/cognitive/src/features/providers/cerebras/cerebras.config.ts b/packages/cognitive/src/features/providers/cerebras/cerebras.config.ts
index 665e691..b7a3d18 100644
--- a/packages/cognitive/src/features/providers/cerebras/cerebras.config.ts
+++ b/packages/cognitive/src/features/providers/cerebras/cerebras.config.ts
@@ -6,6 +6,62 @@ export const CEREBRAS_CONFIG: ProviderConfig = {
   description: 'Cerebras AI models for high-performance language processing',
   defaultModel: 'gpt-oss-120b',
   models: [
+    {
+      id: 'qwen-3-235b-a22b-instruct-2507',
+      displayName: 'Qwen3 235B A22B Instruct',
+      description:
+        'Qwen3 235B A22B Instruct is a large mixture-of-experts reasoning model from Alibaba with 235B parameters and 22B active per forward pass. It delivers strong performance on reasoning, code generation, and agentic tasks.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: false,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 131_000,
+        maxOutputTokens: 16_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.8,
+        outputCostPer1mTokens: 1.6,
+      },
+      tags: ['preview', 'general-purpose', 'reasoning'],
+      releaseDate: '2025-07-01',
+      lifecycle: 'preview',
+    },
+    {
+      id: 'zai-glm-4.7',
+      displayName: 'Z.ai GLM 4.7',
+      description:
+        'Z.ai GLM 4.7 is a large-scale language model from Zhipu AI with 355B parameters, delivering strong performance across reasoning, coding, and multilingual tasks.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: false,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 131_000,
+        maxOutputTokens: 16_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.39,
+        outputCostPer1mTokens: 1.75,
+      },
+      tags: ['preview', 'general-purpose', 'reasoning'],
+      releaseDate: '2025-08-01',
+      lifecycle: 'preview',
+    },
     {
       id: 'gpt-oss-120b',
       displayName: 'GPT-OSS 120B (Preview)',
diff --git a/packages/cognitive/src/features/providers/fireworks-ai/fireworks.config.ts b/packages/cognitive/src/features/providers/fireworks-ai/fireworks.config.ts
index c2a2d42..fdf0101 100644
--- a/packages/cognitive/src/features/providers/fireworks-ai/fireworks.config.ts
+++ b/packages/cognitive/src/features/providers/fireworks-ai/fireworks.config.ts
@@ -6,6 +6,90 @@ export const FIREWORKS_CONFIG: ProviderConfig = {
   description: 'Fireworks AI provides fast inference for various language models',
   defaultModel: 'accounts/fireworks/models/llama-v3p1-70b-instruct',
   models: [
+    {
+      id: 'deepseek-v3p2',
+      internalModelId: 'accounts/fireworks/models/deepseek-v3p2',
+      displayName: 'DeepSeek V3.2',
+      description:
+        'DeepSeek V3.2 is the latest iteration of the DeepSeek V3 series, a 685B-parameter mixture-of-experts model with improved reasoning, instruction following, and coding performance.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: false,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: false,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 163_840,
+        maxOutputTokens: 16_384,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.56,
+        outputCostPer1mTokens: 1.68,
+      },
+      tags: ['recommended', 'general-purpose', 'coding'],
+      lifecycle: 'production',
+    },
+    {
+      id: 'deepseek-v3p1',
+      internalModelId: 'accounts/fireworks/models/deepseek-v3p1',
+      displayName: 'DeepSeek V3.1',
+      description:
+        'DeepSeek V3.1 is an updated version of the DeepSeek V3 series with enhanced reasoning and instruction-following capabilities.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: false,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: false,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 163_840,
+        maxOutputTokens: 16_384,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.56,
+        outputCostPer1mTokens: 1.68,
+      },
+      tags: ['general-purpose', 'coding'],
+      lifecycle: 'production',
+    },
+    {
+      id: 'kimi-k2-instruct-0905',
+      internalModelId: 'accounts/fireworks/models/kimi-k2-instruct-0905',
+      displayName: 'Kimi K2 Instruct 0905',
+      description:
+        'Kimi K2 Instruct 0905 from Moonshot AI is an updated version optimized for agentic tasks, complex reasoning, and coding with a large 262k context window.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: false,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 262_144,
+        maxOutputTokens: 16_384,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.6,
+        outputCostPer1mTokens: 2.5,
+      },
+      tags: ['general-purpose', 'reasoning', 'coding'],
+      lifecycle: 'production',
+    },
     {
       id: 'gpt-oss-20b',
       internalModelId: 'accounts/fireworks/models/gpt-oss-20b',
@@ -109,12 +193,12 @@ export const FIREWORKS_CONFIG: ProviderConfig = {
         supportsTemperature: true,
       },
       limits: {
-        maxInputTokens: 160_000,
+        maxInputTokens: 163_840,
         maxOutputTokens: 16_384,
       },
       cost: {
-        inputCostPer1mTokens: 0.9,
-        outputCostPer1mTokens: 0.9,
+        inputCostPer1mTokens: 0.56,
+        outputCostPer1mTokens: 1.68,
       },
       tags: ['recommended', 'general-purpose'],
       lifecycle: 'production',
diff --git a/packages/cognitive/src/features/providers/google-ai/google-ai.config.ts b/packages/cognitive/src/features/providers/google-ai/google-ai.config.ts
index c5d8e21..350b4fe 100644
--- a/packages/cognitive/src/features/providers/google-ai/google-ai.config.ts
+++ b/packages/cognitive/src/features/providers/google-ai/google-ai.config.ts
@@ -6,6 +6,64 @@ export const GOOGLE_AI_CONFIG: ProviderConfig = {
   description: 'Gemini models from Google with multimodal capabilities',
   defaultModel: 'gemini-2.5-flash',
   models: [
+    {
+      id: 'gemini-3.1-pro',
+      internalModelId: 'gemini-3.1-pro-preview',
+      displayName: 'Gemini 3.1 Pro',
+      description:
+        "Google's most advanced model with multimodal understanding, agentic capabilities, and state-of-the-art reasoning. Features tiered pricing for prompts over 200k tokens.",
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 1_048_576,
+        maxOutputTokens: 65_536,
+      },
+      cost: {
+        inputCostPer1mTokens: 2,
+        outputCostPer1mTokens: 12,
+      },
+      tags: ['reasoning', 'agents', 'general-purpose', 'vision'],
+      releaseDate: '2026-01-15',
+      lifecycle: 'preview',
+    },
+    {
+      id: 'gemini-3.1-flash-lite',
+      internalModelId: 'gemini-3.1-flash-lite-preview',
+      displayName: 'Gemini 3.1 Flash Lite',
+      description:
+        "Google's most cost-effective Gemini 3.1 model, optimized for high-volume and latency-sensitive use cases with strong reasoning capabilities.",
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 1_048_576,
+        maxOutputTokens: 65_536,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.25,
+        outputCostPer1mTokens: 1.5,
+      },
+      tags: ['low-cost', 'reasoning', 'general-purpose', 'vision'],
+      releaseDate: '2026-01-15',
+      lifecycle: 'preview',
+    },
     {
       id: 'gemini-3-pro',
       internalModelId: 'gemini-3-pro-preview',
@@ -31,9 +89,11 @@ export const GOOGLE_AI_CONFIG: ProviderConfig = {
         inputCostPer1mTokens: 2,
         outputCostPer1mTokens: 12,
       },
-      tags: ['reasoning', 'agents', 'general-purpose', 'vision'],
+      tags: ['deprecated', 'reasoning', 'agents', 'general-purpose', 'vision'],
       releaseDate: '2025-11-18',
-      lifecycle: 'preview',
+      lifecycle: 'discontinued',
+      discontinuedDate: '2026-03-09',
+      replacementModels: ['gemini-3.1-pro'],
     },
     {
       id: 'gemini-3-flash',
@@ -119,6 +179,34 @@ export const GOOGLE_AI_CONFIG: ProviderConfig = {
       releaseDate: '2025-01-15',
       lifecycle: 'production',
     },
+    {
+      id: 'gemini-2.5-flash-lite',
+      displayName: 'Gemini 2.5 Flash Lite',
+      description:
+        "Google's most cost-effective stable Gemini model, offering fast inference and solid reasoning capabilities for high-volume applications.",
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 1_048_576,
+        maxOutputTokens: 65_536,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.1,
+        outputCostPer1mTokens: 0.4,
+      },
+      tags: ['low-cost', 'reasoning', 'general-purpose', 'vision'],
+      releaseDate: '2025-06-01',
+      lifecycle: 'production',
+    },
     {
       id: 'gemini-2.0-flash',
       displayName: 'Gemini 2.0 Flash',
@@ -144,9 +232,11 @@ export const GOOGLE_AI_CONFIG: ProviderConfig = {
         inputCostPer1mTokens: 0.1,
         outputCostPer1mTokens: 0.4,
       },
-      tags: ['low-cost', 'general-purpose', 'vision'],
+      tags: ['deprecated', 'low-cost', 'general-purpose', 'vision'],
       releaseDate: '2024-12-11',
-      lifecycle: 'production',
+      lifecycle: 'deprecated',
+      deprecationDate: '2026-01-01',
+      replacementModels: ['gemini-2.5-flash'],
     },
   ],
 }
diff --git a/packages/cognitive/src/features/providers/groq/groq.config.ts b/packages/cognitive/src/features/providers/groq/groq.config.ts
index 0048857..d2f8dbe 100644
--- a/packages/cognitive/src/features/providers/groq/groq.config.ts
+++ b/packages/cognitive/src/features/providers/groq/groq.config.ts
@@ -6,6 +6,93 @@ export const GROQ_CONFIG: ProviderConfig = {
   description: 'Groq provides fast inference for OpenAI GPT OSS models',
   defaultModel: 'llama-3.3-70b-versatile',
   models: [
+    {
+      id: 'llama-4-scout-17b-16e-instruct',
+      displayName: 'Llama 4 Scout 17B (Preview)',
+      internalModelId: 'meta-llama/llama-4-scout-17b-16e-instruct',
+      description:
+        'Llama 4 Scout 17B is a mixture-of-experts model from Meta with 16 experts per forward pass. Supports multimodal input (text and image) and multilingual output across 12 languages.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsSystemMessages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsStreaming: true,
+        supportsReasoning: false,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 131_000,
+        maxOutputTokens: 32_768,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.11,
+        outputCostPer1mTokens: 0.34,
+      },
+      tags: ['preview', 'general-purpose', 'vision'],
+      releaseDate: '2025-04-14',
+      lifecycle: 'production',
+    },
+    {
+      id: 'kimi-k2-instruct-0905',
+      displayName: 'Kimi K2 Instruct 0905 (Preview)',
+      internalModelId: 'moonshotai/kimi-k2-instruct-0905',
+      description:
+        'Kimi K2 Instruct 0905 is an updated version of the Kimi K2 model from Moonshot AI, optimized for agentic and reasoning tasks with strong coding capabilities.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: false,
+        supportsSystemMessages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 131_000,
+        maxOutputTokens: 32_768,
+      },
+      cost: {
+        inputCostPer1mTokens: 1,
+        outputCostPer1mTokens: 3,
+      },
+      tags: ['preview', 'general-purpose', 'reasoning', 'coding'],
+      releaseDate: '2025-09-05',
+      lifecycle: 'production',
+    },
+    {
+      id: 'qwen3-32b',
+      displayName: 'Qwen3 32B (Preview)',
+      internalModelId: 'qwen/qwen3-32b',
+      description:
+        'Qwen3 32B is a world-class reasoning model from Alibaba with strong performance on code generation, tool calling, and advanced reasoning tasks.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: false,
+        supportsSystemMessages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 131_000,
+        maxOutputTokens: 32_768,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.29,
+        outputCostPer1mTokens: 0.59,
+      },
+      tags: ['preview', 'general-purpose', 'reasoning', 'coding'],
+      releaseDate: '2025-05-01',
+      lifecycle: 'production',
+    },
     {
       id: 'gpt-oss-20b',
       displayName: 'GPT-OSS 20B (Preview)',
@@ -25,11 +112,11 @@ export const GROQ_CONFIG: ProviderConfig = {
       },
       limits: {
         maxInputTokens: 131_000,
-        maxOutputTokens: 32_000,
+        maxOutputTokens: 65_536,
       },
       cost: {
-        inputCostPer1mTokens: 0.1,
-        outputCostPer1mTokens: 0.5,
+        inputCostPer1mTokens: 0.075,
+        outputCostPer1mTokens: 0.3,
       },
       tags: ['preview', 'general-purpose', 'reasoning', 'low-cost'],
       releaseDate: '2024-12-01',
@@ -54,11 +141,11 @@ export const GROQ_CONFIG: ProviderConfig = {
       },
       limits: {
         maxInputTokens: 131_000,
-        maxOutputTokens: 32_000,
+        maxOutputTokens: 65_536,
       },
       cost: {
         inputCostPer1mTokens: 0.15,
-        outputCostPer1mTokens: 0.75,
+        outputCostPer1mTokens: 0.6,
       },
       tags: ['preview', 'general-purpose', 'reasoning'],
       releaseDate: '2024-12-01',
@@ -111,7 +198,7 @@ export const GROQ_CONFIG: ProviderConfig = {
         supportsTemperature: true,
       },
       limits: {
-        maxInputTokens: 128_000,
+        maxInputTokens: 131_000,
         maxOutputTokens: 32_768,
       },
       cost: {
@@ -263,8 +350,8 @@ export const GROQ_CONFIG: ProviderConfig = {
         supportsTemperature: true,
       },
       limits: {
-        maxInputTokens: 128_000,
-        maxOutputTokens: 8192,
+        maxInputTokens: 131_000,
+        maxOutputTokens: 131_072,
       },
       cost: {
         inputCostPer1mTokens: 0.05,
diff --git a/packages/cognitive/src/features/providers/openrouter/openrouter.config.ts b/packages/cognitive/src/features/providers/openrouter/openrouter.config.ts
index 1bb7e9a..6452821 100644
--- a/packages/cognitive/src/features/providers/openrouter/openrouter.config.ts
+++ b/packages/cognitive/src/features/providers/openrouter/openrouter.config.ts
@@ -25,14 +25,42 @@ export const OPENROUTER_CONFIG: ProviderConfig = {
       },
       limits: {
         maxInputTokens: 131_000,
-        maxOutputTokens: 32_000,
+        maxOutputTokens: 65_536,
       },
       cost: {
         inputCostPer1mTokens: 0.15,
-        outputCostPer1mTokens: 0.75,
+        outputCostPer1mTokens: 0.6,
       },
       tags: ['preview', 'general-purpose', 'reasoning'],
       lifecycle: 'production',
     },
+    {
+      id: 'gpt-oss-20b',
+      displayName: 'GPT-OSS 20B (Preview)',
+      provider: 'openrouter',
+      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.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: false,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: false,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 131_000,
+        maxOutputTokens: 65_536,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.075,
+        outputCostPer1mTokens: 0.3,
+      },
+      tags: ['preview', 'general-purpose', 'reasoning', 'low-cost'],
+      lifecycle: 'production',
+    },
   ],
 }
diff --git a/packages/cognitive/src/features/providers/xai/xai.config.ts b/packages/cognitive/src/features/providers/xai/xai.config.ts
index 20c3161..0103200 100644
--- a/packages/cognitive/src/features/providers/xai/xai.config.ts
+++ b/packages/cognitive/src/features/providers/xai/xai.config.ts
@@ -4,8 +4,138 @@ export const XAI_CONFIG: ProviderConfig = {
   id: 'xai',
   name: 'xAI',
   description: 'xAI Grok models',
-  defaultModel: 'grok-4-fast-non-reasoning',
+  defaultModel: 'grok-4.20-0309-non-reasoning',
   models: [
+    {
+      id: 'grok-4.20-0309-reasoning',
+      displayName: 'Grok 4.20 (Reasoning)',
+      description: "xAI's flagship Grok 4.20 model with advanced reasoning capabilities and 2M token context window.",
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: false,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 2_000_000,
+        maxOutputTokens: 128_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 2,
+        outputCostPer1mTokens: 6,
+      },
+      tags: ['recommended', 'reasoning', 'general-purpose', 'vision'],
+      lifecycle: 'production',
+    },
+    {
+      id: 'grok-4.20-0309-non-reasoning',
+      displayName: 'Grok 4.20 (Non-Reasoning)',
+      description: "xAI's flagship Grok 4.20 model optimized for fast general-purpose tasks with 2M token context window.",
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: false,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 2_000_000,
+        maxOutputTokens: 128_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 2,
+        outputCostPer1mTokens: 6,
+      },
+      tags: ['recommended', 'general-purpose', 'vision'],
+      lifecycle: 'production',
+    },
+    {
+      id: 'grok-4.20-multi-agent-0309',
+      displayName: 'Grok 4.20 Multi-Agent',
+      description: "xAI's Grok 4.20 model optimized for multi-agent workflows with 2M token context window.",
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: false,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 2_000_000,
+        maxOutputTokens: 128_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 2,
+        outputCostPer1mTokens: 6,
+      },
+      tags: ['agents', 'reasoning', 'general-purpose', 'vision'],
+      lifecycle: 'production',
+    },
+    {
+      id: 'grok-4-1-fast-reasoning',
+      displayName: 'Grok 4.1 Fast (Reasoning)',
+      description: 'Fast Grok 4.1 model with reasoning capabilities and 2M token context window.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: true,
+        supportsStopSequence: false,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 2_000_000,
+        maxOutputTokens: 128_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.2,
+        outputCostPer1mTokens: 0.5,
+      },
+      tags: ['reasoning', 'general-purpose', 'low-cost'],
+      lifecycle: 'production',
+    },
+    {
+      id: 'grok-4-1-fast-non-reasoning',
+      displayName: 'Grok 4.1 Fast (Non-Reasoning)',
+      description: 'Fast, cost-effective Grok 4.1 model for non-reasoning tasks with 2M token context window.',
+      capabilities: {
+        supportsText: true,
+        supportsImages: true,
+        supportsTools: true,
+        supportsJsonMode: true,
+        supportsSystemMessages: true,
+        supportsStreaming: true,
+        supportsReasoning: false,
+        supportsStopSequence: true,
+        supportsTemperature: true,
+      },
+      limits: {
+        maxInputTokens: 2_000_000,
+        maxOutputTokens: 128_000,
+      },
+      cost: {
+        inputCostPer1mTokens: 0.2,
+        outputCostPer1mTokens: 0.5,
+      },
+      tags: ['low-cost', 'general-purpose'],
+      lifecycle: 'production',
+    },
     {
       id: 'grok-code-fast-1',
       displayName: 'Grok Code Fast 1',