Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work
https://integrate.api.nvidia.com/v1 moonshotai/kimi-k3 | Benchmark | Score | Metric | Date |
|---|---|---|---|
| DeepSWE | 67.5 | resolve rate | 2026-07-16 |
| Terminal-Bench | 88.3 | accuracy | 2026-07-16 |
| FrontierSWE | 81.2 | dominance score | 2026-07-16 |
| Program Bench | 77.8 | score | 2026-07-16 |
Kimi K3 is listed for chat workloads and supports a 1.0M context window.
Check NVIDIA NIM's current endpoint status before integrating this listing. Use the alternatives table if the provider listing is unavailable.
We only found this Kimi K3 listing on NVIDIA NIM in the current catalog. Use the related models section for nearby alternatives.
| Provider | Model listing | Access | Context | API | Limits |
|---|---|---|---|---|---|
| | Kimi K3 | Check provider | 1.0M | Native | Varies |
| Model | Provider | Context | Access |
|---|---|---|---|
| 01-ai/yi-large | NVIDIA NIM | 131K | Free tier |
| adept/fuyu-8b | NVIDIA NIM | 131K | Free tier |
| ai21labs/jamba-1.5-large-instruct | NVIDIA NIM | 131K | Free tier |
| aisingapore/sea-lion-7b-instruct | NVIDIA NIM | 131K | Free tier |
| bigcode/starcoder2-15b | NVIDIA NIM | 131K | Free tier |
Kimi K3 is tagged for chat in this catalog and works with OpenAI-compatible client libraries.
Kimi K3 appears in the free model catalog for NVIDIA NIM, but its current endpoint availability should be confirmed with the provider before use.
The model ID shown in this catalog is moonshotai/kimi-k3.
The listed context window is 1.0M tokens with up to 131K output tokens.
Free Kimi K3 API.
For API keys, setup steps, and provider-level limits, see the NVIDIA NIM provider page.