Claude Code, running on GPT.

Pick GPT from your ChatGPT plan in /model, next to Claude. GLM, Kimi, DeepSeek, Gemini and local models work the same way.

brew install baoengineer/tap/byom

~/myappGPT-5.6-Sol

❯ /model

Select model

Switch between Claude models. Your pick becomes the default for new sessions.

↑/↓ to switch model, ←/→ for effort

$ byom run openai/gpt-5.6-sol

Use the plans you already pay for.

Sixteen providers are built in. Anything that speaks the Anthropic or OpenAI API works too, including the model server on your own machine.

~/myappbyom models

$ byom modelsMODELCONTEXT$/M IN/OUTSTATUSopenai/gpt-6-astra272kplanreadyopenai/gpt-5.6-sol272kplanreadyzai/glm-5.3200kplanreadykimi/k3256kplanreadyminimax/MiniMax-M31000kplancappeddeepseek/deepseek-v4-pro128k0.50/2.00readygemini/gemini-flash-latest1000k0.30/2.50readygroq/llama-3.3-70b-versatile128k0.59/0.79readyopenrouter/qwen/qwen3-coder262k0.22/0.95readymistral/devstral-latest128k-no-keyollama/qwen3-coder32klocalreadyclaude-opus-5-51000kplanready

Claude Code running through byom. The model servers in this recording are local mocks.

Put a team of models on one task.

Claude leads. GPT reviews while GLM writes the fix and Kimi writes the tests, each on its own plan, and they message each other as they go.

Claudelead, splits the work

GPT-5.6-Solreviews the diffChatGPT plan
GLM-5.3implements the fixZ.ai
Kimi K3writes the testsKimi
qwen3-coderscans the logsOllama, local

Tasksshared by the team

✓Review the diff in src/pool.rsGPT-5.6-Sol
↳GPT-5.6-Sol → GLM-5.3: release() returns the socket before resetmessage
●Reset before returning the connectionGLM-5.3
○Test concurrent releaseKimi K3
✓Scan logs for double-release errorsqwen3-coder

This runs on Claude Code's agent teams, which are still experimental. Turn them on with byom config set teams true. Without teams, Claude can still hand one task at a time to any model as a subagent.

Sign in, then run.

~/myapp

$ byom login
$ byom