Provider reliability · tracked daily
Which provider is most reliable for each model?
Uptime and latency, per provider, per model - measured directly from OpenRouter's own endpoint data. The uptime figure is OpenRouter's own trailing-30-minute metric, captured once daily by our sync (not a live/real-time feed) - a daily snapshot of that window, not this instant. A model's benchmark score doesn't tell you whether you can actually depend on getting a response; this does.
Which host is most reliable overall?
DeepInfra leads among hosts serving 3+ tracked models, at 100.0% median uptime across 34 models. Inverted from the table above: instead of "which host is best for this model," this ranks each host by measured reliability across every model it serves.
| # | Provider | Models served | Median uptime | Avg uptime | Peak throughput |
|---|---|---|---|---|---|
| 01 | DeepInfra | 34 | 100.0% | 96.1% | 238 tok/s |
| 02 | Azure | 29 | 100.0% | 99.9% | 166 tok/s |
| 03 | 29 | 100.0% | 93.7% | 181 tok/s | |
| 04 | Amazon Bedrock | 28 | 100.0% | 99.7% | 202 tok/s |
| 05 | Novita | 27 | 100.0% | 99.9% | 213 tok/s |
| 06 | GMICloud | 18 | 100.0% | 98.2% | 120 tok/s |
| 07 | Parasail | 18 | 100.0% | 96.4% | 169 tok/s |
| 08 | Venice | 18 | 100.0% | 94.8% | 310 tok/s |
| 09 | CoreWeave | 15 | 100.0% | 100.0% | 269 tok/s |
| 10 | Phala | 15 | 100.0% | 96.7% | 196 tok/s |
| 11 | Together | 15 | 100.0% | 99.5% | 193 tok/s |
| 12 | AtlasCloud | 14 | 100.0% | 99.1% | 132 tok/s |
| 13 | Io Net | 14 | 100.0% | 99.0% | 146 tok/s |
| 14 | OpenAI | 14 | 100.0% | 96.6% | 131 tok/s |
| 15 | SiliconFlow | 14 | 100.0% | 99.9% | 88 tok/s |
| 16 | Alibaba | 13 | 100.0% | 90.1% | 116 tok/s |
| 17 | Anthropic | 13 | 100.0% | 100.0% | 119 tok/s |
| 18 | Fireworks | 13 | 100.0% | 66.7% | 173 tok/s |
| 19 | StreamLake | 13 | 100.0% | 87.0% | 96 tok/s |
| 20 | DigitalOcean | 12 | 100.0% | 99.7% | 199 tok/s |
| 21 | Google AI Studio | 11 | 100.0% | 99.9% | 178 tok/s |
| 22 | Nebius | 11 | 100.0% | 96.9% | 302 tok/s |
| 23 | BaseTen | 10 | 100.0% | 98.9% | 240 tok/s |
| 24 | Claude Platform on AWS | 10 | 100.0% | 100.0% | 189 tok/s |
| 25 | Crusoe | 10 | 100.0% | 100.0% | 245 tok/s |
| 26 | DekaLLM | 10 | 100.0% | 100.0% | 126 tok/s |
| 27 | Relace | 10 | 100.0% | 100.0% | 153 tok/s |
| 28 | Wafer | 10 | 100.0% | 100.0% | 213 tok/s |
| 29 | AkashML | 9 | 100.0% | 99.9% | 131 tok/s |
| 30 | InferenceNet | 9 | 100.0% | 99.6% | 168 tok/s |
| 31 | Makora | 9 | 100.0% | 99.9% | 143 tok/s |
| 32 | Morph | 9 | 100.0% | 99.0% | 107 tok/s |
| 33 | Cloudflare | 8 | 100.0% | 92.8% | 116 tok/s |
| 34 | Darkbloom | 8 | 100.0% | 100.0% | 99 tok/s |
| 35 | PrimeIntellect | 8 | 100.0% | 100.0% | 148 tok/s |
| 36 | Reka | 8 | 100.0% | 99.1% | 108 tok/s |
| 37 | Sail Research | 8 | 100.0% | 100.0% | 96 tok/s |
| 38 | Baidu | 7 | 100.0% | 100.0% | 91 tok/s |
| 39 | Modal | 7 | 100.0% | 100.0% | 163 tok/s |
| 40 | ModelRun | 7 | 100.0% | 98.6% | 219 tok/s |
| 41 | NextBit | 6 | 100.0% | 99.8% | 73 tok/s |
| 42 | SambaNova | 6 | 100.0% | 100.0% | 280 tok/s |
| 43 | Decart | 5 | 100.0% | 78.0% | 140 tok/s |
| 44 | Friendli | 5 | 100.0% | 100.0% | 132 tok/s |
| 45 | Cohere | 4 | 100.0% | 100.0% | 167 tok/s |
| 46 | Ionstream | 4 | 100.0% | 99.6% | 124 tok/s |
| 47 | Mancer 2 | 4 | 100.0% | 100.0% | 79 tok/s |
| 48 | Meta | 4 | 100.0% | 100.0% | 126 tok/s |
| 49 | Mistral | 4 | 100.0% | 100.0% | 130 tok/s |
| 50 | Tenstorrent | 4 | 100.0% | 99.2% | 117 tok/s |
| 51 | Xiaomi | 4 | 100.0% | 99.9% | 33 tok/s |
| 52 | Z.AI | 4 | 100.0% | 100.0% | 66 tok/s |
| 53 | Cerebras | 3 | 100.0% | 66.7% | 835 tok/s |
| 54 | Chutes | 3 | 100.0% | 100.0% | 31 tok/s |
| 55 | Inceptron | 3 | 100.0% | 100.0% | 147 tok/s |
| 56 | Minimax | 3 | 100.0% | 99.5% | 126 tok/s |
| 57 | Near AI | 3 | 100.0% | 100.0% | 75 tok/s |
| 58 | Tencent | 3 | 100.0% | 100.0% | 58 tok/s |
| 59 | xAI | 3 | 100.0% | 99.9% | 79 tok/s |
| 60 | Ambient | 2 | 100.0% | 100.0% | 98 tok/s |
| 61 | Moonshot AI | 2 | 100.0% | 100.0% | 32 tok/s |
| 62 | Nex AGI | 2 | 100.0% | 100.0% | 94 tok/s |
| 63 | Poolside | 2 | 100.0% | 100.0% | 114 tok/s |
| 64 | PrimaLabs | 2 | 100.0% | 100.0% | 109 tok/s |
| 65 | StepFun | 2 | 100.0% | 100.0% | 103 tok/s |
| 66 | Upstage | 2 | 100.0% | 100.0% | 44 tok/s |
| 67 | Featherless | 1 | 100.0% | 100.0% | 25 tok/s |
| 68 | Krea | 1 | 100.0% | 100.0% | 96 tok/s |
| 69 | ParetoAlpha | 1 | 100.0% | 100.0% | 23 tok/s |
| 70 | Particle AI | 1 | 100.0% | 100.0% | 124 tok/s |
| 71 | Stealth | 4 | 100.0% | 100.0% | 58 tok/s |
| 72 | DeepSeek | 4 | 100.0% | 99.6% | 78 tok/s |
| 73 | Groq | 3 | 99.9% | 99.9% | 453 tok/s |
| 74 | OpenInference | 4 | 99.4% | 94.7% | 103 tok/s |
| 75 | Mara | 4 | 98.1% | 98.1% | 186 tok/s |