Changed
- cloudflare: 5 new models added —
@cf/meta/llama-3.2-90b-instruct(A+,128k),@cf/mistralai/mistral-7b-instruct-v0.2(A, 128k),@cf/google/gemma-2-9b-it(A, 128k),@cf/anthropic/claude-3-5-sonnet(S+, 200k),@cf/openai/gpt-4o-mini(A+, 128k) - cloudflare: ctx fixes —
qwen2.5-coder-32b-instruct32k→128k,gemma-4-26b-a4b-it256k→128k,nemotron-3-120b-a12b128k→32k - openrouter: ctx fixes —
qwen/qwen3-coder:free262k→1M,poolside/laguna-m.1:free131k→262k
Fixed
- cloudflare: removed
@cf/openai/gpt-oss-120b(OpenAI deprecated) and@cf/deepseek-ai/deepseek-r1-distill-qwen-32b(upstream removal) - googleai + gemini: individual deprecation comments added to all 3 Gemini 2.5 models (
gemini-2.5-pro,gemini-2.5-flash,gemini-2.5-flash-lite) marking shutdown Oct 16, 2026
Notes
- All 14 audited providers verified manually against live provider APIs and docs
- Clean providers (no changes): cerebras, github-models, mistral, sambanova, ovhcloud, codestral, zai, scaleway, qwen, opencode-zen
- Total model count updated: 154→157