Command Code

Available models

Pass any of these ids to -m, --model, or pick them in-session with /model. Run cmd --list-models for the live, copy-pasteable list; that command is the source of truth if this table ever drifts. The default model currently is moonshotai/Kimi-K2.5.

cmd --model claude-sonnet-4-6 # interactive session on Sonnet 4.6 cmd -p "summarize this repo" --model gpt-5.5 # one-shot on GPT-5.5 cmd --model kimi-k2.5 # short form: name after the "/" cmd --list-models # print every id you can pass

--model matching is case-insensitive and accepts either the full id or just the name after the /, so moonshotai/Kimi-K2.5, moonshotai/kimi-k2.5, and kimi-k2.5 all resolve to the same model. An unknown id is rejected up front (pointing you at cmd --list-models) rather than failing later with a cryptic provider error. Each model id below links to its profile page; models that don't have a dedicated page yet link to the models overview.

Anthropic

Model idNameBest forCapabilities
claude-sonnet-4-6Claude Sonnet 4.6best combo of speed & intelligence (recommended)text, vision
claude-fable-5Claude Fable 5most capable for demanding reasoning & long-horizon agentstext, vision
claude-opus-4-8Claude Opus 4.8most capable for complex reasoning & agentic codingtext, vision
claude-opus-4-7Claude Opus 4.7most intelligent for agents and codingtext, vision
claude-haiku-4-5Claude Haiku 4.5fastest & most compact, great for quick taskstext, vision

OpenAI

Model idNameBest forCapabilities
gpt-5.5GPT-5.5latest frontier model for general complex worktext, vision
gpt-5.4GPT-5.4frontier model for general complex worktext, vision
gpt-5.3-codexGPT-5.3 Codexfrontier coding modeltext, vision
gpt-5.4-miniGPT-5.4 Minifast, cost-effective model for everyday taskstext, vision

Google

Model idNameBest forCapabilities
google/gemini-3.5-flashGemini 3.5 FlashPro-level coding proficiency, parallel agentic executiontext, vision
google/gemini-3.1-flash-liteGemini 3.1 Flash Litehigh-volume workhorse model with implicit cachingtext, vision

Open Source

Model idNameBest forCapabilities
moonshotai/Kimi-K2.6Kimi K2.6long-horizon coding with visiontext, vision
moonshotai/Kimi-K2.5 (default)Kimi K2.5multimodal frontend codingtext, vision
zai-org/GLM-5.1GLM-5.1long-horizon autonomous coding agenttext
zai-org/GLM-5GLM-5multi-mode thinking & long-range planningtext
MiniMaxAI/MiniMax-M3MiniMax M3frontier coding, agents & native multimodalitytext, vision
MiniMaxAI/MiniMax-M2.7MiniMax M2.7end-to-end software engineering agenttext
MiniMaxAI/MiniMax-M2.5MiniMax M2.5cross-platform full-stack agentic devtext
deepseek/deepseek-v4-proDeepSeek V4 Prohybrid-attention long-context reasoningtext
deepseek/deepseek-v4-flashDeepSeek V4 Flashfast hybrid-attention reasoningtext
Qwen/Qwen3.6-Max-PreviewQwen 3.6 Max Previewvibe coding & efficient agent executiontext
Qwen/Qwen3.6-PlusQwen 3.6 Plusagentic coding & reasoningtext, vision
Qwen/Qwen3.7-MaxQwen 3.7 Maxfrontier coding & long-horizon agent executiontext
Qwen/Qwen3.7-PlusQwen 3.7 Plusagentic coding & reasoning at lower costtext, vision
stepfun/Step-3.7-FlashStep 3.7 Flashmultimodal sparse-MoE reasoningtext, vision
stepfun/Step-3.5-FlashStep 3.5 Flashfast sparse-MoE agentic reasoningtext
xiaomi/mimo-v2.5-proMiMo V2.5 Prohigh-capability long-context agentic codingtext
xiaomi/mimo-v2.5MiMo V2.5efficient long-context agentic codingtext, vision
nvidia/nemotron-3-ultra-550b-a55bNemotron 3 Ultraopen reasoning model for long-horizon autonomous agentstext

Next steps