How do I use GPT models like GPT-5.6 Sol in Claude Code?
OpenAI's API does not speak Anthropic's Messages format, so Claude Code needs an endpoint that translates: set ANTHROPIC_BASE_URL to https://api.stdcmpt.com, a Standard Compute key in ANTHROPIC_AUTH_TOKEN and ANTHROPIC_MODEL to openai/gpt-5.6-sol, and Claude Code runs on GPT-5.6 Sol on a plan from $19 a month. When the plan's compute budget hits 100%, new requests get HTTP 402.
# GPT-5.6 Sol for the session, Terra for subagents, GPT-6 Luna for background calls
# No /v1 here: Claude Code adds /v1/messages itself.
export ANTHROPIC_BASE_URL=https://api.stdcmpt.com
export ANTHROPIC_AUTH_TOKEN=YOUR_API_KEY
export ANTHROPIC_MODEL=openai/gpt-5.6-sol
export ANTHROPIC_DEFAULT_OPUS_MODEL=openai/gpt-5.6-sol
export ANTHROPIC_DEFAULT_SONNET_MODEL=openai/gpt-5.6-terra
export ANTHROPIC_DEFAULT_HAIKU_MODEL=openai/gpt-6-luna
export CLAUDE_CODE_SUBAGENT_MODEL=openai/gpt-5.6-terra
claude
Replace YOUR_API_KEY with a key from the dashboard. In the low-cost tab, /model opus still reaches Sol for a hard step.
Sol, Terra or Luna in Claude Code
The three GPT tiers differ by about 40 times in daily cost, so the slot each one takes matters more than which one you pin.
| Model ID | Typical day | Busy day | Good slot |
|---|---|---|---|
| openai/gpt-5.6-sol | $8.21 | $83 | Main session or opus, for hard steps |
| openai/gpt-5.6-terra | $4.19 | $43 | Subagents and sonnet |
| openai/gpt-6-luna | $0.21 | $2.08 | Haiku and background calls, or all day on a budget |
OpenAI list prices per million tokens, prompts up to 272K: GPT-5.6 Sol $4 input, $0.40 cached, $20 output; Terra $2, $0.20, $12; GPT-6 Luna $0.10, $0.01, $0.50. Longer prompts pay OpenAI's long-context rates. Source. A typical day is 8.5M tokens and a busy one 86.7M, 87.2% of input as cache reads: the median and 90th-percentile coding-agent day in our Agent Burn Index, measured on OpenCode, Cline, Kilo and similar agents, September 13 to October 12, 2026. Priced at the list rates above; your own days will differ.
What works with GPT in Claude Code
| Claude Code feature | On GPT through Standard Compute |
|---|---|
| Tools and MCP servers | Work, as OpenAI function calls |
| Pasted screenshots | Work on all three |
| Thinking toggle | No effect; pick each model's reasoning effort on the Models page |
| Context | About 1M tokens here, with requests capped near 850K; Claude Code still assumes 200K unless told otherwise |
Claude Code's WebSearch tool is not available on this endpoint for any model.
The catch
Sol all day is costly on any route: $164 for 20 typical days at OpenAI's list price, and the $20 Starter budget covers about 2 such days. If you already pay for ChatGPT, OpenAI's own Codex CLI is the native way to use these models, and LiteLLM or claude-code-router can translate for Claude Code on your machine if you want to pay OpenAI per token.
Where Standard Compute fits
Nothing to host: the translation runs on our side, Claude and GPT sit on one key, and /model switches between them mid-session. Luna-heavy setups fit Starter; Sol-heavy ones need Standard ($89 a month) or above.
FAQ
Can I use my ChatGPT Plus or Pro plan in Claude Code?
No. A ChatGPT plan only covers OpenAI's own apps and Codex sign-in. Claude Code needs an API key on an Anthropic-format endpoint.
Is GPT-6 Astra available too?
Yes, as openai/gpt-6-astra. At $10 input and $50 output per million tokens it is the most expensive OpenAI model here, so it fits the opus slot for occasional hard steps rather than the main session.
Does GPT handle Claude Code's tools well?
Claude Code's prompts and tools are written for Claude, and Anthropic does not support non-Claude models behind a gateway. Tool calls go through, but try a real task before moving your daily work.
Does my Claude Pro or Max subscription pay for this?
No. While ANTHROPIC_AUTH_TOKEN holds a gateway key, Claude Code sends that key instead of your claude.ai login, so the subscription's limits and billing do not apply. Remove the variables to go back to it.