Day 0 Claude Opus 5 release

[mateo-berri]

Claude Opus 5 is now live on LiteLLM

Call it through the LiteLLM AI Gateway using the same OpenAI-compatible request you already use, across Anthropic, Azure, Vertex AI, and Bedrock, with spend tracking, rate limits, fallbacks, and logging in one place. No SDK changes, no per-provider rewrites.

What's new:

• Much better coding agents: stronger at large refactors, debugging, multi-file features, and finishing end-to-end work without leaving stubs
• Better self-verification: it checks and iterates on its own work more proactively
• Better vision and artifacts: improved screenshots → website, charts and documents, spreadsheets, and slide decks
• Thinking is on by default
• Pricing unchanged from Opus 4.8: $5 / MTok input and $25 / MTok output, with prompt caching at $0.50 / MTok (read) and $6.25 / MTok (5-minute write). The cache minimum drops from 1,024 tokens to 512, pulling shorter system prompts into the read rate. On Bedrock, the us., eu., au., and jp. inference profiles carry the usual 10% regional premium while global. stays at base price; LiteLLM tracks every variant automatically.
• Fast mode at 2x, not 6x: speed: "fast" on the Anthropic provider buys roughly 2.5x faster output at $10 / MTok input and $50 / MTok output, on /v1/chat/completions and /v1/messages today with Responses API support coming soon. LiteLLM sets the fast-mode-2026-02-01 beta header and prices the premium for you.

Worth checking before you switch off Opus 4.8: Priority Tier and web fetch are both gone on Opus 5, and the cybersecurity classifiers can decline a request outright with stop_reason: "refusal"; handle that stop reason or route around it with LiteLLM fallbacks. Anthropic's Opus 5 migration guide has the full compatibility checklist.

Get started, no redeploy needed:

Claude Opus 5 is already in the LiteLLM model cost map. Reload your pricing in the UI under Models + Endpoints → Price Data → Reload Price Data (or POST /reload/model_cost_map as an admin). That enables claude-opus-5 across Anthropic, Azure, Vertex AI, and Bedrock with cost tracking.

model_list:
  - model_name: claude-opus-5
    litellm_params:
      model: anthropic/claude-opus-5
      api_key: os.environ/ANTHROPIC_API_KEY

Running with LITELLM_LOCAL_MODEL_COST_MAP=true? Pull ghcr.io/berriai/litellm:v1.95.0-dev.3 or later for the bundled metadata; that image cuts later today.

Read the full guide → Day 0 Support: Claude Opus 5

Mateo Wang
LiteLLM AI Team

---

Please use this space to voice thoughts and opinions, or any issues you find