AI API providers
21 providers| Serves | ||||
|---|---|---|---|---|
 Venice AIInference host | 109193 | 131 | 2 | 20.3M |
 Prime InferenceInference host | 122 | 122 | — | 172K |
 CortecsInference host | 1021 | 103 | 3 | 17K |
 Deep InfraInference host | 862 | 88 | 24 | 502K |
 Novita AIInference host | 8012 | 83 | 5 | 303K |
 NvidiaInference host | 63195 | 78 | — | — |
 NEAR AI CloudInference host | 5511111 | 60 | — | 19K |
 SiliconFlowInference host | 454244 | 59 | 15 | 1.5M |
| SiliconFlowInference host | 48 | 48 | 2 | 309K |
 Together AIInference host | 211182 | 42 | 3 | 998K |
 PerplexityInference host | 28 | 28 | — | — |
 Charm HyperInference host | 21 | 21 | 1 | — |
 MistralOfficial API | 11721 | 21 | 1 | — |
 BoundlessInference host | 18 | 18 | 1 | 6K |
 TensorXInference host | 14 | 14 | — | 40K |
 Fireworks AIInference host | 10 | 10 | 1 | 622K |
| GroqInference host | 343 | 10 | — | 3.6M |
| BasetenInference host | 9 | 9 | 2 | 443K |
 NeuralwattInference host | 9 | 9 | — | 120K |
 UnoRouterInference host | 81 | 9 | 3 | 336K |
 CerebrasInference host | 2 | 2 | — | 802K |
Visits: an estimate of each provider's website for September 2026, that one month rather than an average; all countries, desktop and mobile, subdomains included. The line shows up to the last 12 months. It shows how well known a provider is, not its API volume.
Free LLM API providers
Providers that charge $0 per token for LLMs: 72 free models across 7 providers. Free access usually comes with rate limits.
LLM API pricing comparison by provider
The most searched LLMs and what each provider charges for them. Green is the cheapest in the row.
| Model | Maker |  AIHubMix | OpenRouter | Vercel AI Gateway | AI/ML API |  CometAPI | DevPass (LLM Gateway) |  Requesty |
|---|---|---|---|---|---|---|---|---|
 Gemini 4 Argon | $10From | $10 | $10From | $10 | $10 | $8 | $20From | $10From |
| Kimi K3 | $13.89 | $14.20 | $10.80From | $15From | $15 | $12 | $15From | $15From |
| GPT-6 Astra | $50From | $50From | $50From | $50From | $50 | $40From | $50From | $50From |
 Muse Wonder | — | $22 | $20From | $20 | $21.16 | $16 | $20From | $20From |
| Claude Opus 5.5 | $20From | $20 | $20From | $20From | $20 | $16From | $20From | $20From |
| GPT-6.1 Sol | $10From | $10From | $10From | $10From | $10 | $8From | $10From | $10From |
 GLM-5.3 | $4.40 | $3.94 | $1.40From | $4From | $4.40 | $3.53 | $3From | $3.42From |
| GPT-6 Sol | $10From | $10From | $10From | $10From | $10 | $8From | $10From | $10From |
 Qwen3.8 2.4T A95B | $6From | $6 | $6From | $6 | $6.61 | $4.80 | $6From | $6From |
| Grok 4.7 | $6From | $6From | $6From | $6 | $6 | $4.80From | $6From | $6From |
Cheapest in the row2× the cheapest or moreOutput price per 1M tokens
Types of AI API providers
Where you can buy model access, and how each kind differs.
- Model makers17 providersThe companies that make the models and sell them through their own API.
- Aggregators29 providersOne key for models from many makers; they pass your calls on to the providers.
- Clouds8 providersBig clouds hosting models on their platform, billed to your cloud account.
- Inference hosts21 providersHosts running open-weight models on their own GPUs, often faster or cheaper.
- Media generation3 providersPlatforms for image, video and audio generation models.
AI API provider rankings
Who is cheapest most often, who lists the most models, and whose site gets the most visits.