OpenCode

OpenCode speaks Chat Completions through an OpenAI-compatible provider passed in for the run.

curl -fsSL https://opencode.ai/install | bash      # install
superfluid launch opencode --model unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_M

What launch gives it

given	what for
OPENCODE_CONFIG_CONTENT	a config with a superfluid provider (@ai-sdk/openai-compatible, the server's /v1 and key) and the served model with the server's context limit, selected as model

OpenCode merges that config over your own, so your providers and settings stay as they are.