API providers
Who actually serves the models, what they charge for them, and how fast they answer. Prices and endpoint telemetry are ingested from OpenRouter — we do not run our own provider benchmarks.
72 hosts serving tracked models
72 of 72 providers
| Provider | Models served | Mean output t/s | Mean TTFT | Lowest blended $/M |
|---|---|---|---|---|
Novita
|
68 | 40 | 2.3s | $0.000 |
DeepInfra
|
65 | 48 | 1.3s | $0.022 |
OpenAI
· first party |
62 | 74 | 2.3s | $0.069 |
Azure
· first party |
46 | 57 | 3.1s | $0.14 |
Alibaba
· first party |
45 | 63 | 0.9s | $0.11 |
Google
· first party |
45 | 87 | 1.9s | $0.12 |
| Si SiliconFlow | 35 | 37 | 1.6s | $0.075 |
| Pa Parasail | 34 | 57 | 0.9s | $0.060 |
| Ve Venice | 30 | 55 | 1.1s | $0.11 |
| At AtlasCloud | 27 | 52 | 1.4s | $0.17 |
| AB Amazon Bedrock · first party | 24 | 73 | 1.4s | $0.061 |
Google AI Studio
|
23 | 129 | 2.3s | $0.000 |
| Co CoreWeave | 20 | 96 | 0.4s | $0.055 |
| St StreamLake | 20 | 51 | 1.5s | $0.084 |
| Ph Phala | 19 | 42 | 2.2s | $0.055 |
Together
|
18 | 63 | 0.6s | $0.075 |
Mistral
· first party |
17 | 55 | 0.4s | $0.10 |
| Di DigitalOcean | 16 | 29 | 0.9s | $0.14 |
Anthropic
· first party |
14 | 54 | 2.0s | $2.00 |
| Cl Cloudflare | 13 | 51 | 0.7s | $0.041 |
| Gm GMICloud | 13 | 47 | 2.3s | $0.100 |
| Ne Nebius | 13 | 63 | 0.5s | $0.10 |
Z.AI
· first party |
11 | 36 | 5.7s | $0.42 |
| Ne NextBit | 10 | 42 | 0.7s | $0.060 |
| Ch Chutes | 9 | 36 | 1.8s | $0.18 |
Groq
|
9 | 257 | 0.3s | $0.058 |
Fireworks
|
8 | 97 | 1.2s | $0.13 |
Minimax
· first party |
8 | 48 | 1.0s | $0.42 |
Baidu
· first party |
7 | 48 | 1.1s | $0.11 |
| Cr Crusoe | 7 | 61 | 0.4s | $0.087 |
| Mo Morph | 7 | 423 | 0.9s | $0.17 |
Nvidia
· first party |
7 | 67 | 1.1s | $0.000 |
BaseTen
|
6 | 158 | 0.7s | $0.20 |
| Fr Friendli | 6 | 81 | 1.1s | $0.20 |
| Po Poolside | 6 | 65 | 1.1s | $0.000 |
SambaNova
|
6 | 107 | 1.4s | $0.34 |
| Ak AkashML | 5 | 68 | 1.4s | $0.12 |
Cohere
· first party |
5 | 51 | 0.4s | $0.000 |
| Io Ionstream | 5 | 51 | 1.0s | $0.17 |
| M2 Mancer 2 | 5 | 27 | 0.9s | $0.17 |
Perplexity
· first party |
5 | 72 | 21.4s | $1.00 |
xAI
· first party |
5 | 188 | 1.6s | $1.25 |
| Ai AionLabs | 4 | 49 | 1.0s | $0.88 |
| In Inceptron | 4 | 60 | 0.6s | $0.34 |
| IN Io Net | 4 | 76 | 1.2s | $0.27 |
| Ma Mara | 4 | 150 | 1.4s | $0.30 |
Moonshot AI
· first party |
4 | 56 | 2.1s | $1.20 |
| Se Seed · first party | 4 | 51 | 1.2s | $0.13 |
| Am Ambient | 3 | 57 | 0.9s | $0.17 |
Cerebras
|
3 | 376 | 0.3s | $0.45 |
| Wa Wafer | 3 | 68 | 4.1s | $1.50 |
| Da Darkbloom | 2 | 24 | 4.1s | $0.000 |
| De Decart | 2 | 97 | 1.0s | $1.35 |
DeepSeek
· first party |
2 | 58 | 1.0s | $0.17 |
| In Inflection | 2 | — | — | $4.38 |
| Mo ModelRun | 2 | 74 | 0.3s | $0.30 |
| NA Nex AGI | 2 | 124 | 0.9s | $0.044 |
| Re Reka · first party | 2 | 9 | 1.2s | $0.10 |
| Re Relace | 2 | 14 | 0.5s | $0.95 |
| Xi Xiaomi · first party | 2 | 48 | 2.0s | $0.17 |
| Ai AI21 · first party | 1 | 17 | 0.6s | $3.50 |
| AA Arcee AI | 1 | 46 | 0.7s | $0.39 |
| In Inception | 1 | 371 | 0.8s | $0.38 |
Meta
· first party |
1 | 120 | 5.5s | $2.00 |
| Op OpenInference | 1 | 51 | 1.1s | $0.16 |
| Pe Perceptron | 1 | 31 | 0.9s | $0.49 |
| SR Sail Research | 1 | 27 | 1.1s | $1.72 |
| SA Sakana AI | 1 | 18 | 6.5s | $11.25 |
| St StepFun | 1 | 34 | 4.1s | $0.44 |
Tencent
· first party |
1 | 48 | 3.0s | $0.23 |
Upstage
· first party |
1 | 5 | 0.6s | $0.26 |
| CP Claude Platform on AWS | 1 | — | — | $10.00 |
How to read this. A host appears here once we have seen it serve at least one tracked model.
Mean output t/s and mean TTFT average that host's measurements across every model it serves, so a host
that only serves small models will look faster than one carrying frontier models — compare hosts on a single model
page for a like-for-like reading. Lowest blended $/M is the cheapest blended rate that host publishes for any
model, at a 3:1 input:output blend. Figures are ingested from OpenRouter endpoint listings and telemetry, with
metadata enriched from models.dev. See methodology and
attribution.