feat(sdk): refresh OpenRouter default model list

[mmabrouk]

What

Refresh the OpenRouter default model list in sdks/python/agenta/sdk/utils/assets.py to the ~20 most-used OpenRouter models as of 2026-07-01. The previous list still led with 2025-era models (Claude 3.5/3.7 Sonnet, GPT-4o, Gemini 2.x, Mixtral, Llama-3-70B, Grok-4) that have fallen off OpenRouter's usage leaderboard.

This one list feeds three consumers, so every id is kept valid: the litellm completion path, the Pi harness model picker (capabilities.py _pi_models() reads supported_llm_models["openrouter"]), and the frontend picker. The openrouter/ prefix format is preserved on every entry.

Ranking source

OpenRouter's public usage rankings (https://openrouter.ai/rankings and https://openrouter.ai/models), cross-referenced against three mid-2026 token-volume / agent-spend leaderboards (officechai June 2026, digitalapplied April 2026, codesota agent-usage snapshot 2026-06-23). Date of ranking capture: 2026-07-01. Selection blends raw usage with vendor breadth for a practical default picker.

Final list (20)

# Anthropic
openrouter/anthropic/claude-opus-4.8
openrouter/anthropic/claude-opus-4.7
openrouter/anthropic/claude-sonnet-4.6
openrouter/anthropic/claude-sonnet-4.5
# DeepSeek
openrouter/deepseek/deepseek-v4-flash
openrouter/deepseek/deepseek-v4-pro
openrouter/deepseek/deepseek-v3.2
# Google
openrouter/google/gemini-3.5-flash
openrouter/google/gemini-3-flash-preview
openrouter/google/gemini-3.1-pro-preview
# MiniMax
openrouter/minimax/minimax-m3
# MoonshotAI
openrouter/moonshotai/kimi-k2.6
# Nvidia
openrouter/nvidia/nemotron-3-super-120b-a12b
# OpenAI
openrouter/openai/gpt-5.5
openrouter/openai/gpt-5.4
# Qwen
openrouter/qwen/qwen3.7-max
# Tencent
openrouter/tencent/hy3-preview
# Xiaomi
openrouter/xiaomi/mimo-v2.5-pro
# xAI
openrouter/x-ai/grok-4.3
# Z.ai
openrouter/z-ai/glm-5

Added (18 new ids)

claude-opus-4.8, claude-opus-4.7, claude-sonnet-4.6, deepseek-v4-flash, deepseek-v4-pro, gemini-3.5-flash, gemini-3-flash-preview, gemini-3.1-pro-preview, minimax-m3, kimi-k2.6, nemotron-3-super-120b-a12b, gpt-5.5, gpt-5.4, qwen3.7-max, hy3-preview, mimo-v2.5-pro, grok-4.3, glm-5.

Removed (21 stale ids)

claude-opus-4.5, claude-opus-4.1, claude-haiku-4.5, claude-3.7-sonnet, claude-3.5-sonnet, deepseek-chat, deepseek-r1, deepseek-r1-0528, gemini-2.5-pro, gemini-2.5-flash, gemini-2.0-flash-001, llama-3-70b-instruct, mistral-large, mistral-small-3.2-24b-instruct, mixtral-8x22b-instruct, gpt-4o, gpt-4.1, gpt-5, qwen-2.5-coder-32b-instruct, qwen3-235b-a22b-2507, grok-4 (upgraded to grok-4.3).

Kept from the old list: claude-sonnet-4.5, deepseek-v3.2. Meta-Llama and Mistral are dropped as vendors — neither appears in any 2026 OpenRouter usage top-20; Chinese OSS models (DeepSeek, MiniMax, Moonshot, Xiaomi, Tencent, Z.ai) and the frontier families displaced them.

Two selections are vendor-breadth/continuity picks rather than strict top-20 usage cites: x-ai/grok-4.3 (keeps xAI represented, upgrading the old grok-4) and one of the two Gemini flashes; all others are directly grounded in the ranking sources.

Pi vendored-catalog cross-check

Each id (minus the openrouter/ prefix) must be a valid key in Pi's static OpenRouter catalog to be settable by the Pi harness. Checked against @earendil-works/pi-ai@0.79.4 (dist/models.generated.js, openrouter block).

Result: all 20 ids are present in Pi's vendored catalog. None are missing.

Verification

• from agenta.sdk.utils.assets import supported_llm_models; len(supported_llm_models["openrouter"]) → 20.
• ruff format + ruff check --fix on assets.py: clean.
• oss/tests/pytest/unit/agents/connections/test_capabilities.py: 12 passed.

Note on PR base

Base is big-agents, not main. This GitButler workspace targets big-agents, so a main-based PR would show ~1,634 unrelated files (the whole big-agents stack). assets.py is identical on main and big-agents, so this change applies cleanly to either; big-agents is used to keep the diff to exactly one file for review.

https://claude.ai/code/session_01HCMtsTWnCdh8fPEzGrda6C

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
agenta-documentation	Ready	Preview, Comment	Jul 1, 2026 6:35pm

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 608edefb-bb0f-45bb-a524-d57dfea50059

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch openrouter-top-models

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}