chore(cognitive): update AI model catalog

[github-actions]

Model Update Summary

Updated 18 models across 4 providers (anthropic, google-ai, groq, xai). No changes needed for openai, cerebras, openrouter, fireworks-ai.

---

Anthropic

New Models Added (5)

Model ID	Change	Details
claude-opus-4-6	ADDED	Latest flagship model. $5/$25 per 1M tokens, 200K input / 128K output, lifecycle: production
claude-sonnet-4-6	ADDED	Latest Sonnet model. $3/$15 per 1M tokens, 200K input / 64K output, lifecycle: production
claude-opus-4-5-20251101	ADDED	Claude Opus 4.5. $5/$25 per 1M tokens, 200K input / 64K output, lifecycle: production
claude-opus-4-1-20250805	ADDED	Claude Opus 4.1. $15/$75 per 1M tokens, 200K input / 32K output, lifecycle: production
claude-opus-4-20250514	ADDED	Claude Opus 4. $15/$75 per 1M tokens, 200K input / 32K output, lifecycle: production

Updated Models (1)

Model ID	Field	Old Value	New Value
claude-3-haiku-20240307	lifecycle	production	deprecated
claude-3-haiku-20240307	deprecationDate	(none)	2026-04-19
claude-3-haiku-20240307	replacementModels	(none)	['claude-haiku-4-5-20251001']

Other Config Changes

Field	Old Value	New Value
defaultModel	claude-sonnet-4-5-20250929	claude-sonnet-4-6

Source: https://docs.anthropic.com/en/docs/about-claude/models

---

Google AI

New Models Added (3)

Model ID	Change	Details
gemini-3.1-pro	ADDED	Latest Google flagship preview. internalModelId: gemini-3.1-pro-preview. $2/$12 per 1M tokens, 1M input / 65K output, lifecycle: preview
gemini-3.1-flash-lite	ADDED	Frontier-class efficiency preview. internalModelId: gemini-3.1-flash-lite-preview. $0.25/$1.5 per 1M tokens, 1M input / 65K output, lifecycle: preview
gemini-2.5-flash-lite	ADDED	Fastest/cheapest 2.5-family model. $0.10/$0.40 per 1M tokens, 1M input / 65K output, lifecycle: production

Updated Models (2)

Model ID	Field	Old Value	New Value
gemini-3-pro	lifecycle	preview	deprecated
gemini-3-pro	tags	[...'reasoning'...]	added 'deprecated'
gemini-3-pro	replacementModels	(none)	['gemini-3.1-pro']
gemini-2.0-flash	lifecycle	production	deprecated
gemini-2.0-flash	tags	[...'general-purpose'...]	added 'deprecated'
gemini-2.0-flash	replacementModels	(none)	['gemini-2.5-flash']

Source: https://ai.google.dev/gemini-api/docs/models, https://ai.google.dev/gemini-api/docs/pricing

---

Groq

No New Models

Updated Models (4)

Model ID	Field	Old Value	New Value
gpt-oss-20b	inputCostPer1mTokens	0.1	0.075
gpt-oss-20b	outputCostPer1mTokens	0.5	0.3
gpt-oss-20b	maxInputTokens	131_000	131_072
gpt-oss-20b	maxOutputTokens	32_000	65_536
gpt-oss-120b	outputCostPer1mTokens	0.75	0.6
gpt-oss-120b	maxInputTokens	131_000	131_072
gpt-oss-120b	maxOutputTokens	32_000	65_536
llama-3.3-70b-versatile	maxInputTokens	128_000	131_072
llama-3.1-8b-instant	maxInputTokens	128_000	131_072
llama-3.1-8b-instant	maxOutputTokens	8_192	131_072

Source: https://console.groq.com/docs/models

---

xAI

New Models Added (2)

Model ID	Change	Details
grok-4-1-fast-reasoning	ADDED	Grok 4.1 fast model with reasoning. $0.20/$0.50 per 1M tokens, 2M input / 128K output, lifecycle: production
grok-4-1-fast-non-reasoning	ADDED	Grok 4.1 fast model without reasoning. $0.20/$0.50 per 1M tokens, 2M input / 128K output, lifecycle: production

Other Config Changes

Field	Old Value	New Value
defaultModel	grok-4-fast-non-reasoning	grok-4-1-fast-non-reasoning

Source: https://docs.x.ai/docs/models

---

OpenAI

No changes needed. Config already includes the latest models (GPT-5.2, GPT-5.1, GPT-5, GPT-5 Mini, GPT-5 Nano, o4-mini, o3, GPT-4.1 series, o3-mini, o1 series, GPT-4o series). Note: OpenAI docs returned HTTP 403 during fetch; based on existing config state the models appear current.

---

Cerebras

No changes made. Two new preview models were found in the API reference (qwen-3-235b-a22b-instruct-2507 and zai-glm-4.7) but could not be added because pricing (required schema field) was not available in official documentation at the time of this update.

---

OpenRouter

No changes needed. Config intentionally contains only gpt-oss-120b as a pass-through model per existing design. No additional models were added per "only top models" instruction scope.

---

Fireworks AI

No changes needed. Existing models (deepseek-r1-0528, deepseek-v3-0324, llama4-maverick-instruct-basic, llama4-scout-instruct-basic, llama-v3p3-70b-instruct, gpt-oss-120b, gpt-oss-20b) are confirmed current. A newer DeepSeek V3.1 ($0.56/$1.68) was seen on the Fireworks website but could not be added without a confirmed internal model ID.

[allardy]

hey