Open-weight and open-source models via API
Open-weight does not mean you must self-host. On Geek Hub you run DeepSeek, Llama, Mistral, and Kimi with the same DX as GPT/Claude — useful to diversify providers, cost, or weight-related compliance.
Models in this collection
| Model | Type | Context | Price |
|---|---|---|---|
 DeepSeek V4 FlashDeepSeek | Chat | 1M tokens | $0.1526/M in · $0.3052/M out |
| DeepSeek V4 ProDeepSeek | Chat | 1M tokens | $0.4742/M in · $0.9484/M out |
 Llama 3.3 70B InstructMeta | Chat | 128k tokens | $0.6431/M in · $0.8611/M out |
 Mistral SmallMistral | Chat | 128k tokens | $0.109/M in · $0.327/M out |
| Mistral LargeMistral | Chat | 128k tokens | $2.1801/M in · $6.5404/M out |
| Ministral 8BMistral | Chat | 128k tokens | $0.109/M in · $0.109/M out |
 Kimi K3Moonshot (Kimi) | Chat | 1M tokens | $3.2702/M in · $16.3509/M out |
Why these models
DeepSeek pushes price/quality; Llama 3.3 70B is the classic instruct; Mistral covers small→large; Kimi K3 adds 1M context in the open-friendly family.
Use them with Geek Hub
One OpenAI-compatible base URL and API key. Swap any model id from this list without rewriting your client.
Get an API keyFAQ
- Can I download weights and self-host?
- That’s outside Geek Hub. Here you consume the model as an API. If you later self-host, swap the endpoint; the OpenAI-compatible contract helps.
- Are they worse than closed models?
- On many benchmarks they already compete. For coding/RP/reasoning, measure on your eval — don’t assume.
- ZDR / data?
- Check ZDR caps per model on the model page. Policies vary by provider.