NVIDIA NIM
Free OpenAI-compatible API access to a large catalog of hosted open models on NVIDIA's developer platform.
- Requests / min
- 40
- Card required
- No
NVIDIA NIM scores 6.5/10 for free value in LLM. The main constraint: Free developer access: sign up and call many models through one endpoint; free use is governed by a 40 requests-per-minute rate limit (NVIDIA retired the old signup-credit system). Verify current rate limits and model list on the site.
Free value score
6.5 / 10Show the math
Key metrics
Models
38 free · 139+ totalLimits may be stale — last re-verified outside this section's freshness window.
| Model | Family | AA | Access |
|---|---|---|---|
| minimax-m3 | MiniMax | 44.4 | free |
| deepseek-v4-pro | DeepSeek | 44.3 | free |
| kimi-k2.6 | Kimi | 42.8 | free |
| deepseek-v4-flash | DeepSeek | 40.3 | free |
| glm-5.1 | GLM | 40.2 | free |
| minimax-m2.7 | MiniMax | 38.1 | free |
| nemotron-3-ultra-550b-a55b | Nemotron | 37.8 | free |
| qwen3.5-122b-a10b | Qwen | 32.3 | free |
| qwen3.5-397b-a17b | Qwen | 32 | free |
| mistral-medium-3.5-128b | Mistral | 29.9 | free |
| step-3.7-flash | Step (StepFun) | 29.7 | free |
| gpt-oss-120b | GPT (OpenAI open) | 23.8 | free |
| mistral-small-4-119b-2603 | Mistral | 20.8 | free |
| mistral-large-3-675b-instruct-2512 | Mistral | 16.2 | free |
| gpt-oss-20b | GPT (OpenAI open) | 14.9 | free |
| nemotron-3-nano-omni-30b-a3b-reasoning | Nemotron | 14.9 | free |
| llama-4-maverick-17b-128e-instruct | Llama | 14.3 | free |
| nemotron-3-nano-30b-a3b | Nemotron | 7.4 | free |
| diffusiongemma-26b-a4b-it | Gemma | — | free |
| gemma-2-2b-it | Gemma | — | free |
| gemma-3n-e2b-it | Gemma | — | free |
| gemma-3n-e4b-it | Gemma | — | free |
| gemma-4-31b-it | Gemma | — | free |
| llama-3.1-70b-instruct | Llama | — | free |
| llama-3.1-8b-instruct | Llama | — | free |
| llama-3.1-nemotron-nano-8b-v1 | Nemotron (Llama) | — | free |
| llama-3.2-90b-vision-instruct | Llama | — | free |
| llama-3.3-70b-instruct | Llama | — | free |
| llama-3.3-nemotron-super-49b-v1.5 | Nemotron (Llama) | — | free |
| ministral-14b-instruct-2512 | Mistral | — | free |
| mistral-nemotron | Mistral/Nemotron | — | free |
| mixtral-8x7b-instruct-v0.1 | Mistral | — | free |
| nemotron-3-super-120b-a12b | Nemotron | — | free |
| nvidia-nemotron-nano-9b-v2 | Nemotron | — | free |
| paligemma | Gemma | — | free |
| qwen3-next-80b-a3b-instruct | Qwen | — | free |
| seed-oss-36b-instruct | Seed (ByteDance) | — | free |
| step-3.5-flash | Step (StepFun) | — | free |
| qwen-image | Qwen | — | paid |
| qwen-image-edit | Qwen | — | paid |
Showing first 38 of a larger catalog.
What you get
Large hosted model catalog
OpenAI-compatible: integrate.api.nvidia.com
Free rate-limited access (~40 RPM)
No card required to start
Gotchas
- NVIDIA developer account required
Users must register a free NVIDIA developer account; no credit card is needed but identity signup is required.
Sources
-
"We no longer use a credit-based system for build.nvidia.com. As you noted, this has been replaced by rate limits for trial usage. The rate limits vary for each model, and we do not publish those."
https://forums.developer.nvidia.com/t/request-more-4-000-credits-option-on-build-nvidia-com/344567 -
"When you create the account and get the API key you are using the NIM as a trial. You don't need to do anything else first. The trial period is not limited by a time period."
https://forums.developer.nvidia.com/t/request-more-4-000-credits-option-on-build-nvidia-com/344567 -
"Current Limit: 40 RPM ... The current limit frequently causes 429 "Too Many Requests" errors"
https://forums.developer.nvidia.com/t/request-for-nvidia-nim-api-rate-limit-increase-40-200-rpm/369357 -
"Moonshotai ... Free Endpoint ... kimi-k2.6 ... 1T multimodal MoE for long-horizon coding, agentic tool use, and image/video understanding. ... DeepSeek AI ... Free Endpoint ... deepseek-v4-pro ... Z.ai ... Free Endpoint ... glm-5.1"
https://build.nvidia.com/models
Alternatives
Same category · similar quota
Two stacked packs: 1,000 requests/5h + $48 API credits/week + 2 concurrent requests per model across GLM-5.2 (512k), Kimi-K2.7-Code, MiniMax M3 and other open models; native Anthropic + OpenAI endpoints
GLM-5.2 (512k) + Kimi-K2.7-Code + MiniMax M3 + GLM-4.7-Flash + gpt-oss-120b flat-rate: 500 requests/5h + $24 API credits/week per pack, limits regenerate continuously (no lockouts); native Anthropic + OpenAI endpoints; stack packs to scale linearly
OVHcloud AI Endpoints offers a permanent free anonymous tier — no signup or API key — to call 40+ EU-hosted open-weight models (Llama, Qwen, Mixtral, DeepSeek-R1 distill) via an OpenAI-compatible API.