AI Providers
NVIDIA NIM
NVIDIA NIM provides hosted inference for large open-source models with fast GPU clusters. New accounts receive free API credits, and it has some of the largest models available including DeepSeek R1 671B and Qwen3 Coder 480B.
Getting your API key
- 1Visit integrate.api.nvidia.com and sign in with your NVIDIA account.
- 2Go to API Keys and generate a new key.
- 3Copy the key (starts with
nvapi-). - 4Paste it in Capsul under Settings → API Keys → NVIDIA NIM.
Available models in Capsul
| Model | ID |
|---|---|
| Nemotron Ultra 253B | nvidia/llama-3.1-nemotron-ultra-253b-v1 |
| Kimi K2 | moonshotai/kimi-k2 |
| DeepSeek R1 671B | deepseek-ai/deepseek-r1 |
| DeepSeek V3 | deepseek-ai/deepseek-v3-0324 |
| Llama 4 Maverick | meta/llama-4-maverick-17b-128e-instruct |
| GLM-4.1V 9B Thinking | thudm/glm-4.1v-9b-thinking |
| Qwen3 Coder 480B | qwen/qwen3-coder-480b-a35b-instruct |
| QwQ 32B | qwen/qwq-32b |
Free credits
New NVIDIA accounts receive a generous amount of free credits. Usage and remaining credits are visible in the NVIDIA developer portal. Because Capsul uses your own key, you are billed directly by NVIDIA — Capsul never charges you for AI usage.