Inference provider ranking
# providers
100 providers serving 134 open-weights models, aggregated from live per-model offers. verified = models the provider's own catalog/price API confirms · top-10 = how many of the current overall top 10 they host · cheapest on = models where they have the best blended price · vs best $ = median premium over each model's cheapest offer (1.00× = usually the cheapest). Click a column to sort. Machine-readable: providers.json.
| # | provider | models | verified | top-10 | cheapest on | vs best $ | tps (med) | ttft (med) |
|---|---|---|---|---|---|---|---|---|
| 1 | DeepInfra | 87 | 48 | 4 | 32 | 1.19× | 34 | 540ms |
| 2 | Novita | 74 | 54 | 2 | 15 | 1.42× | 45 | 1001ms |
| 3 | SiliconFlow | 56 | 50 | 2 | 6 | 1.94× | — | — |
| 4 | Together | 48 | — | 2 | 3 | 2.59× | — | — |
| 5 | Featherless | 48 | — | 2 | 1 | 1.00× | — | — |
| 6 | Venice | 44 | 41 | 2 | 1 | 1.82× | — | — |
| 7 | Nebius | 42 | 24 | 2 | 5 | 1.50× | — | — |
| 8 | Amazon Bedrock | 33 | 27 | — | 7 | 1.38× | — | — |
| 9 | AtlasCloud | 31 | 31 | 2 | — | 2.50× | — | — |
| 10 | Kyma | 30 | — | 1 | — | 1.89× | — | — |
| 11 | Parasail | 30 | — | 1 | — | 2.44× | — | — |
| 12 | Alibaba | 29 | — | 1 | 4 | 3.02× | — | — |
| 13 | GMICloud | 27 | — | 2 | 7 | 1.51× | — | — |
| 14 | StreamLake | 27 | — | 2 | 3 | 1.88× | — | — |
| 15 | Wandb | 26 | — | 2 | 2 | 2.09× | — | — |
| 16 | BaseTen | 26 | — | 2 | 1 | 1.95× | — | — |
| 17 | DigitalOcean | 21 | — | 1 | — | 2.46× | — | — |
| 18 | Phala | 21 | 21 | — | — | 4.27× | — | — |
| 19 | Aihubmix | 20 | — | 2 | 4 | 1.55× | — | — |
| 20 | CoreWeave | 20 | — | 1 | 2 | 2.26× | — | — |
| 21 | Fireworks | 19 | — | 2 | 1 | 2.50× | — | — |
| 22 | Azure | 19 | 1 | 1 | 3 | 2.50× | — | — |
| 23 | Cloudflare | 19 | — | 1 | 1 | 5.05× | — | — |
| 24 | Opencode Go | 17 | — | 2 | 2 | 1.56× | — | — |
| 25 | Crusoe | 17 | — | 1 | — | 3.70× | — | — |
| 26 | Chutes | 15 | 13 | — | 1 | 2.51× | — | — |
| 27 | Tencent | 14 | — | 2 | 4 | 2.48× | — | — |
| 28 | Inferenceapis | 14 | — | 2 | — | 2.96× | — | — |
| 29 | Aiand | 14 | — | 1 | 1 | 4.35× | — | — |
| 30 | Scaleway | 13 | — | — | — | 7.72× | 129 | 450ms |
| 31 | Baidu | 12 | — | 1 | — | 2.48× | — | — |
| 32 | Mistral | 12 | 3 | — | 5 | 1.41× | — | — |
| 33 | PrimeIntellect | 12 | — | — | 5 | 1.45× | — | — |
| 34 | Nscale | 12 | — | — | 4 | 1.17× | 81 | 400ms |
| 35 | Hyperbolic | 12 | 5 | — | 3 | 2.18× | — | — |
| 36 | Vultr | 11 | — | 1 | 5 | 1.57× | — | — |
| 37 | Relace | 11 | — | 1 | 3 | 1.43× | — | — |
| 38 | Z.AI | 11 | — | — | 2 | 1.45× | 73 | 1528ms |
| 39 | Sail | 11 | — | — | 1 | 2.55× | — | — |
| 40 | Clf Ai Gateway | 10 | — | 1 | 4 | 2.14× | — | — |
| 41 | Qwen Ai Platform | 10 | — | 1 | 2 | 2.14× | — | — |
| 42 | Qwencloud | 10 | — | 1 | 2 | 2.14× | — | — |
| 43 | DekaLLM | 10 | — | — | 1 | 2.16× | — | — |
| 44 | Wafer | 10 | — | — | 1 | 1.83× | — | — |
| 45 | SambaNova | 9 | 5 | 1 | — | 5.73× | — | — |
| 46 | Darkbloom | 9 | — | — | 3 | 2.11× | — | — |
| 47 | OVHcloud | 9 | 7 | — | 1 | 4.84× | 40 | 473ms |
| 48 | DeepSeek | 8 | — | 1 | 1 | 1.45× | — | — |
| 49 | Sail Research | 8 | — | — | 2 | 1.83× | — | — |
| 50 | Makora | 8 | — | — | 1 | 2.37× | — | — |
| 51 | AkashML | 8 | — | — | — | 2.85× | — | — |
| 52 | Minimax | 7 | — | 1 | 2 | 1.40× | — | — |
| 53 | Modal | 7 | — | — | 1 | 5.12× | — | — |
| 54 | Replicate | 7 | — | — | 1 | 5.53× | — | — |
| 55 | Friendliai | 7 | — | — | — | 3.21× | — | — |
| 56 | Reka | 7 | — | — | — | 3.38× | — | — |
| 57 | Morph | 6 | — | — | 3 | 1.00× | — | — |
| 58 | Bedrock Mantle | 6 | — | — | — | 2.55× | — | — |
| 59 | Cohere | 6 | — | — | — | 2.82× | — | — |
| 60 | 6 | — | — | — | 4.82× | — | — | |
| 61 | Groq | 6 | 3 | — | — | 6.69× | — | — |
| 62 | Inceptron | 6 | — | — | — | 3.39× | — | — |
| 63 | InferenceNet | 6 | — | — | — | 2.52× | — | — |
| 64 | Mancer 2 | 6 | — | — | — | 3.30× | — | — |
| 65 | Tensormesh | 6 | — | — | — | 2.09× | — | — |
| 66 | Libertai | 5 | — | — | 1 | 2.02× | — | — |
| 67 | Friendli | 5 | — | — | — | 4.81× | — | — |
| 68 | Io Net | 5 | — | — | — | 2.22× | — | — |
| 69 | NextBit | 5 | — | — | — | 1.50× | — | — |
| 70 | Nvidia | 5 | — | — | — | — | — | — |
| 71 | Mara | 4 | — | 1 | — | 6.16× | — | — |
| 72 | ModelRun | 4 | — | 1 | — | 3.91× | — | — |
| 73 | Gmi | 4 | — | — | 1 | 1.19× | — | — |
| 74 | Tenstorrent | 4 | — | — | 1 | 2.21× | — | — |
| 75 | Moonshot AI | 4 | — | — | — | 2.50× | — | — |
| 76 | Near AI | 4 | — | — | — | 3.22× | — | — |
| 77 | Scx | 4 | — | — | — | 3.02× | — | — |
| 78 | Xiaomi | 4 | — | — | — | 1.30× | — | — |
| 79 | Zerogpu | 4 | — | — | — | 6.87× | — | — |
| 80 | Google Vertex | 3 | — | — | 1 | 3.59× | — | — |
| 81 | Ionstream | 3 | — | — | — | 2.89× | — | — |
| 82 | OpenInference | 3 | — | — | — | 1.77× | — | — |
| 83 | Nex AGI | 2 | — | — | 2 | 1.00× | — | — |
| 84 | StepFun | 2 | — | — | 2 | 1.00× | — | — |
| 85 | Bigmodel | 2 | — | — | — | 4.40× | — | — |
| 86 | Cerebras | 2 | 1 | — | — | 10.61× | — | — |
| 87 | Coralbricks | 2 | — | — | — | 4.21× | — | — |
| 88 | Particle AI | 2 | — | — | — | 3.20× | — | — |
| 89 | Prism | 2 | — | — | — | 2.52× | — | — |
| 90 | Xiaomi Mimo | 2 | — | — | — | 1.28× | — | — |
| 91 | Lambda | 1 | — | — | 1 | 1.00× | — | — |
| 92 | Liquid | 1 | — | — | 1 | 1.00× | — | — |
| 93 | LongCat | 1 | — | — | 1 | 1.00× | — | — |
| 94 | Ambient | 1 | — | — | — | 2.36× | — | — |
| 95 | Decart | 1 | — | — | — | 3.24× | — | — |
| 96 | Doubleword | 1 | — | — | — | 5.34× | — | — |
| 97 | IBM watsonx | 1 | — | — | — | 8.83× | — | — |
| 98 | ParetoAlpha | 1 | — | — | — | 3.16× | — | — |
| 99 | Refuseless | 1 | — | — | — | 1.14× | — | — |
| 100 | Snowflake | 1 | — | — | — | 2.46× | — | — |
default order: model coverage, then top-10 coverage, then price wins. Prices are blended $/M tokens (3:1 input:output) from each provider's live offers; a model's per-provider offers are on its own page.