# NIM Stats — Fleet Overview

> Real-time status and reliability metrics for free NVIDIA NIM API endpoints. Track throughput, latency, uptime, and congestion across Llama, Mistral, Gemma, Phi, Qwen, DeepSeek, and more models.

- **Fleet state:** Partial degradation
- **Endpoints tracked:** 24
- **Healthy:** 1 · **Busy:** 9 · **Jammed:** 14
- **Average time to first token:** 2175 ms
- **Average throughput:** 22.4 tok/s
- **Last probe:** 2026-09-27T12:32:43.750Z

## What this is

NIM Stats continuously probes every free NVIDIA NIM chat-completion endpoint with a real request and publishes what it measured: time to first token, end-to-end latency, sustained throughput, success rate, and congestion. Numbers below are measurements taken from outside NVIDIA's network, not vendor-published status claims. Use them to pick an endpoint that is working right now instead of discovering it by trial and error.

## Recommended endpoint right now

**nemotron-3-super-120b-a12b** (NVIDIA) — 100% reliability, 502 ms TTFT, 63.6 tok/s, 2% congestion.

Why: Low congestion (2%); High session reliability; Low queue pressure; Excellent uptime (100%); Minimal timeout rate.

## Model fleet

| Model | Provider | Status | Uptime | TTFT (ms) | Throughput (tok/s) | Congestion |
| --- | --- | --- | --- | --- | --- | --- |
| gpt-oss-20b | Other | Busy | 100.00% | 2001 | 16.0 | 6% |
| llama-3.2-11b-vision-instruct | Meta | Busy | 100.00% | 200 | 46.6 | 3% |
| nemotron-3-nano-omni-30b-a3b-reasoning | NVIDIA | Busy | 100.00% | 485 | 61.9 | 5% |
| nemotron-3-super-120b-a12b | NVIDIA | Busy | 100.00% | 502 | 63.6 | 2% |
| nemotron-3.5-content-safety | NVIDIA | Healthy | 100.00% | 189 | 21.3 | 5% |
| riva-translate-4b-instruct-v2 | NVIDIA | Busy | 100.00% | 11655 | 0.8 | 16% |
| kimi-k3 | Other | Busy | 96.67% | 714 | 14.8 | 11% |
| muse-glimmer-30b | Meta | Busy | 96.67% | 2332 | 13.7 | 9% |
| riva-translate-4b-instruct-v1.1 | NVIDIA | Busy | 96.67% | 241 | 28.8 | 4% |
| nemotron-3-ultra-550b-a55b | NVIDIA | Busy | 93.33% | 744 | 36.0 | 13% |
| diffusiongemma-26b-a4b-it | Google | Jammed | 90.00% | 607 | 37.9 | 17% |
| laguna-xs-2.1 | Other | Jammed | 80.00% | 183 | 5.5 | 23% |
| glm-5.3 | Other | Jammed | 76.67% | 5174 | 6.2 | 28% |
| gemma-4-31b-it | Google | Jammed | 70.00% | 1370 | 12.0 | 30% |
| nemotron-3.5-lightning-30b-a3b | NVIDIA | Jammed | 70.00% | 1721 | 18.6 | 27% |
| ising-calibration-1.5-31b | NVIDIA | Jammed | 56.67% | 573 | 16.8 | 36% |
| llama-3.1-nemoguard-8b-topic-control | NVIDIA | Jammed | 56.67% | 149 | 0.0 | 100% |
| llama-3.1-nemotron-safety-guard-8b-v3 | NVIDIA | Jammed | 40.00% | 8822 | 0.8 | 35% |
| mistral-nemotron | Mistral | Jammed | 40.00% | 4862 | 2.5 | 45% |
| glm-5.3-flash | Other | Jammed | 6.67% | 971 | 0.0 | 100% |
| deepseek-v4.1-flash | DeepSeek | Jammed | 0.00% | 0 | 0.0 | 100% |
| llama-3.1-nemoguard-8b-content-safety | NVIDIA | Jammed | 0.00% | 0 | 0.0 | 100% |
| llama-3.2-90b-vision-instruct | Meta | Jammed | 0.00% | 0 | 0.0 | 100% |
| llama-guard-4-12b | Meta | Jammed | 0.00% | 0 | 0.0 | 100% |

Status meanings: **Healthy** — serving normally. **Busy** — serving with elevated latency or congestion. **Jammed** — failing or timing out on probe.

---

## Machine-readable index

- [Fleet overview](https://nimstats.aathil.com) — live status of every tracked endpoint
- [Discover](https://nimstats.aathil.com/discover) — rankings, provider comparison, trends
- [Public status](https://nimstats.aathil.com/status) — plain status summary
- [API reference](https://nimstats.aathil.com/api) — public JSON API, OpenAPI 3.1 at /openapi.json
- [About](https://nimstats.aathil.com/about) — what is measured and how
- [Contact](https://nimstats.aathil.com/contact) — how to reach the operator
- [Privacy](https://nimstats.aathil.com/privacy) — data handling
- [llms.txt](https://nimstats.aathil.com/llms.txt) — agent instructions and when to use this site
- [sitemap.xml](https://nimstats.aathil.com/sitemap.xml) — all indexable URLs

Every page above is available as Markdown at the same URL with `Accept: text/markdown`, or by appending `.md` to the path.

_NIM Stats is an independent project and is not affiliated with NVIDIA Corporation._