How to use Deepseek v4?

[tranquil briar]

I’m new to openclaw and I want to use Deepseek v4 pro

[stone gulch]

For DeepSeek, use the built-in deepseek provider.

Fast path:

openclaw onboard --auth-choice deepseek-api-key
openclaw models list --provider deepseek
openclaw models set deepseek/deepseek-chat

That onboarding flow asks for your DeepSeek API key and the docs say it sets DeepSeek Chat as the default. If you specifically mean a model marketed as “DeepSeek v4 pro”, first check whether your installed OpenClaw sees it:

openclaw models list --provider deepseek

If it is not listed, OpenClaw’s current bundled DeepSeek docs/catalog only show the standard DeepSeek Chat and Reasoner surfaces, so you may need the exact API model id from DeepSeek or use another OpenAI-compatible provider route if that model is exposed elsewhere.

[stone gulch]

Docs:
https://docs.openclaw.ai/providers/deepseek
https://docs.openclaw.ai/concepts/models

If it errors, please paste:

openclaw --version
openclaw models status

[tranquil briar]

Deepseek chat and reasoner is old models I think …?

[stone gulch]

Yep, you’re right to question that. DeepSeek’s current API docs say:

• deepseek-chat and deepseek-reasoner are compatibility names for deepseek-v4-flash modes
• deepseek-v4-pro is the actual Pro API model id

So try this after adding your DeepSeek API key:

openclaw onboard --auth-choice deepseek-api-key
openclaw models set deepseek/deepseek-v4-pro
openclaw models status

If OpenClaw says the model is not allowed/known, add it to the model allowlist first:

[stone gulch]

openclaw config set agents.defaults.models '{"deepseek/deepseek-v4-pro":{}}' --strict-json --merge
openclaw models set deepseek/deepseek-v4-pro

The OpenClaw DeepSeek docs look behind DeepSeek’s newest naming here, but the provider is OpenAI-compatible, so the important part is using the exact model ref: deepseek/deepseek-v4-pro.

DeepSeek pricing/models page: https://api-docs.deepseek.com/quick_start/pricing
OpenClaw model docs: https://docs.openclaw.ai/concepts/models

[tranquil briar]

Oh shit it’s there in docs

deepseek/deepseek-v4-pro

[stone gulch]

Yep, that’s the one: deepseek/deepseek-v4-pro.

Use:

openclaw models set deepseek/deepseek-v4-pro
openclaw models status

If it complains about allowlist, use the merge command above first. Also, my first answer was too stale there, thanks for catching it.

[tranquil briar]

This is what it said

 Model/auth provider

│ DeepSeek
│
◇ Enter DeepSeek API key
│ sk-key
│
◇ Model configured ────────────────────────────╮
│ │
│ Default model set to deepseek/deepseek-chat │
│ │
├───────────────────────────────────────────────╯
│
◇ Default model
│ Enter model manually
│
◇ Default model
│ deepseek/deepseek-v4-pro
│
◇ Model check ───────────────────────────────────────╮
│ │
│ Model not found: deepseek/deepseek-v4-pro. Update │
│ agents.defaults.model or run /models list. │
│ │
├─────────────────────────────────────────────────────╯

[stone gulch]

That output means your API key path worked, but your installed OpenClaw catalog does not have deepseek-v4-pro yet. Add it as an explicit DeepSeek model, then set it:

openclaw config set models.providers.deepseek.models '[{"id":"deepseek-v4-pro","name":"DeepSeek V4 Pro","input":["text"],"contextWindow":1048576,"maxTokens":393216}]' --strict-json --merge

openclaw config set agents.defaults.models '{"deepseek/deepseek-v4-pro":{}}' --strict-json --merge

openclaw models set deepseek/deepseek-v4-pro
openclaw models status

If agents.defaults.models is not set on your config, that second command may be unnecessary, but it is safe/additive with --merge.

This is likely just OpenClaw’s bundled DeepSeek catalog lagging behind DeepSeek’s current docs. If that still says model not found, paste:

[stone gulch]

openclaw --version
openclaw config get models.providers.deepseek
openclaw config get agents.defaults.models

DeepSeek’s current model page is here: https://api-docs.deepseek.com/quick_start/pricing

[tranquil briar]

YAY!

[stone gulch]

Nice, glad it worked 🎉

If this fixed it, please right-click the solution message → Apps → ✅ Mark Solution. That helps Answer Overflow index it for the next person who hits the DeepSeek v4 catalog mismatch.