GPT-6.1 + Claude Code: the catch
GPT-6.1 is an OpenAI model, served over OpenAI's OpenAI-compatible API. Claude Code expects the Anthropic Messages API shape, so, unlike DeepSeek, Kimi or Qwen, which publish Anthropic-compatible endpoints, GPT usually needs a small proxy that translates between the two formats. Once that translator is in place, Claude Code can drive GPT-6.1 like any other endpoint.
The config
What to set
- A translating proxyRun a small proxy that accepts Anthropic-shaped requests and forwards them to OpenAI's GPT-6.1 API.
- ANTHROPIC_BASE_URLPoint Claude Code at your local proxy, not directly at OpenAI.
- ANTHROPIC_AUTH_TOKENYour OpenAI API key (or whatever the proxy expects), billed to your OpenAI account.
- ModelSelect a GPT-6.1 model using the exact name OpenAI's API exposes.
Why people want GPT-6.1 for coding
What GPT-6.1 brings
| Reason | Detail |
|---|---|
| Reasoning | OpenAI's frontier line is strong on multi-step reasoning and tricky debugging |
| Ecosystem | Familiar if your team already lives in the OpenAI stack |
| Another fallback | One more provider to switch to when Claude is capped |
The thing to actually test is multi-step tool calling. A coding agent chains many tool calls per turn, and a model that is great at plain chat can still be uneven at deciding which tool to call next inside Claude Code's loop, especially through a translating proxy, which can subtly reshape tool-call formatting. Run one real agent task end to end (not a one-shot prompt) before trusting GPT-6.1 for long autonomous loops.
The cleaner option: skip the proxy
Because GPT-6.1 is OpenAI-compatible, the friction is entirely Claude Code's Anthropic-only assumption. An agent that speaks OpenAI-compatible endpoints natively doesn't need a translating proxy at all, you add GPT-6.1 as a provider with its base URL and key, and it just works. That removes the extra service and the restart-to-switch dance.
What that looks like with Wuwei
No proxy
OpenAI-compatible providers connect directly, no Anthropic-shape translator to run.
GPT-6.1 in the model list
Pick it next to Claude, DeepSeek, Kimi, Qwen and GLM; switch in a click.
Free, open source, local
MIT-licensed, runs on your machine, reads your real files; you only pay OpenAI's usage.
FAQ
Do I need a proxy for GPT-6.1 in Claude Code?
Usually yes. GPT-6.1 is OpenAI-compatible, not Anthropic-shaped, so Claude Code needs a small translating proxy in front. Agents that speak OpenAI-compatible endpoints natively (like Wuwei) skip the proxy entirely.
Can Claude Code use GPT-6.1?
Yes, with an extra step. Point ANTHROPIC_BASE_URL at a translating proxy, use your OpenAI key as the token, and select a GPT-6.1 model. Or use an agent that supports OpenAI-compatible providers directly.
Is GPT-6.1 good for coding in an agent?
OpenAI's frontier models are strong on reasoning and debugging. As always, test multi-step tool calling on a real task first, proxies can reshape tool-call formatting, so verify before heavy loops.
How do I use GPT-6.1 without running a proxy?
Use an agent that supports OpenAI-compatible providers natively. Wuwei is free and open source, add GPT-6.1 with its base URL and key, no translating proxy, and switch to Claude or other models in a click.