SeaWorkAI models you can use with SeaWork
SeaWork ships with 14 models from 6 model vendors. Claude, OpenAI GPT-5.6, Google Gemini, DeepSeek, Zhipu GLM, and MiniMax are all selectable through SeaWork's built-in gateway — install once and pick any of them, with no separate API key for each model vendor.
Related reading: Claude Fable 5 on your phone · GPT-5.6 with Codex from your phone · Claude Code pricing guide
Anthropic — Claude
claude-fable-5Flagship · Claude 5 familyAnthropic's Claude 5 flagship, redeployed globally on July 1, 2026. Built for long, hard, high-stakes coding — cross-cutting refactors and bugs that beat lesser models. Its premium price earns out on tasks where getting it right in one pass beats a cheaper model that thrashes.
claude-opus-4-8Deep reasoning workhorseAnthropic's Opus-class model — the dependable heavy driver for demanding coding and reasoning when you want strong quality without Fable 5's flagship rate. A solid everyday choice for real repository work in Claude Code.
OpenAI — GPT
gpt-5.6-solFlagship · hardest tasksOpenAI's GPT-5.6 flagship tier, generally available in Codex since July 9, 2026. Reach for it on genuinely hard, open-ended work — a cross-cutting refactor or a bug that's resisted you. A Sol Ultra mode fans work out across coordinating subagents for the biggest tasks.
gpt-5.6-terraBalanced · everyday codingThe mid tier of the GPT-5.6 family: roughly GPT-5.5-class quality at about half the price. The sweet spot for the bulk of everyday coding — feature work, tests, and migrations you already understand.
gpt-5.6-lunaFast & cheap · high volumeThe fastest, cheapest GPT-5.6 tier. Best for high-volume, low-stakes churn — boilerplate, doc passes, mechanical renames, and quick 'explain this file' queries where speed and cost beat the last few points of capability.
gpt-5.5Previous GPT flagshipOpenAI's prior-generation GPT flagship, still a strong general-purpose coding model. A reliable pick when you want proven behavior in Codex without moving to the 5.6 family.
gpt-5.4Established general modelAn earlier GPT generation that remains capable for mainstream coding and reasoning tasks. Useful when you want a stable, well-understood model for routine work.
gpt-5.4-miniCompact · low costThe compact GPT-5.4 variant, tuned for speed and low cost. A good fit for lightweight tasks, quick edits, and high-throughput calls where a smaller model is enough.
Google — Gemini
gemini-3.1-proGoogle flagship · long contextGoogle's Gemini 3.1 Pro — the flagship Gemini tier, strong at reasoning and large-context work. A good option when a task spans a lot of files or documents at once.
gemini-3.5-flashFast & efficientGoogle's Gemini 3.5 Flash — the speed-and-efficiency tier of the Gemini line. Suited to quick, high-volume tasks where latency and cost matter more than peak capability.
DeepSeek
deepseek-v4-proOpen-weight · strong codingDeepSeek's V4 Pro — the higher-capability tier of the V4 line, aimed at strong coding and reasoning. A capable alternative to the frontier US labs, selectable through SeaWork's gateway.
deepseek-v4-flashFast & economicalDeepSeek's V4 Flash — the fast, economical V4 tier for everyday and high-throughput tasks where you want solid quality at lower cost.
Zhipu AI — GLM
glm-5.2Zhipu flagshipZhipu AI's GLM-5.2 — a general-purpose model with solid coding and reasoning, and one of the leading models out of China. Available in SeaWork for teams who prefer or need it.
MiniMax
minimax-m3MiniMax general modelMiniMax's M3 — a general-purpose model covering coding and reasoning tasks. Rounds out SeaWork's provider lineup with another selectable option through the built-in gateway.
How model selection works in SeaWork
You choose an agent provider and a model when you create an agent. For DeepSeek V4, keep the provider set to Claude and select the model ID deepseek-v4-pro or deepseek-v4-flash. SeaWork's built-in gateway supplies the model catalog, so there is no separate DeepSeek API key to manage.
Frequently asked questions
Which AI models can I use with SeaWork?
Fourteen models from six model vendors: Anthropic's Claude Fable 5 and Opus 4.8; OpenAI's GPT-5.6 (Sol, Terra, Luna), GPT-5.5, GPT-5.4, and GPT-5.4-mini; Google's Gemini 3.1 Pro and Gemini 3.5 Flash; DeepSeek V4 Pro and V4 Flash; Zhipu's GLM-5.2; and MiniMax M3. All are selectable through SeaWork's built-in gateway.
Do I need a separate API key for each provider?
No. SeaWork routes every model through its built-in gateway, so you don't sign up for and manage a separate API key from Anthropic, OpenAI, Google, DeepSeek, Zhipu, and MiniMax. Install SeaWork and all 14 models are ready to select — that single-gateway access is the core convenience.
Can I switch models?
Yes. You choose the provider and model when you create an agent, and you can start another agent on a different model at any time. For DeepSeek V4, use the Claude provider with deepseek-v4-pro or deepseek-v4-flash as the model ID.