AI inference hosts
Open-weights models like Llama, Qwen, Gemma or GLM are served by many inference providers at different prices. This page lists every host we track, how many models it serves, and its cheapest option. Prices are base-tier list prices in USD per million tokens.
// hosts
| host | models | cheapest model | in / out $/1M | best price on |
|---|---|---|---|---|
| Amazon Bedrock | 51 | Gemma 4 E2B IT | $0.04 / $0.08 | 32 models |
| Novita AI | 50 | Llama 3.2 1B Instruct (instruct) | $0.02 / $0.02 | 20 models |
| DeepInfra | 47 | Mistral Small 3 | $0.05 / $0.08 | 40 models |
| Azure | 35 | Claude Haiku 5.5 | $0.10 / $0.50 | 30 models |
| Google Vertex AI | 35 | Mistral Small 3.1 24B | $0.10 / $0.30 | 30 models |
| Cloudflare Workers AI | 23 | Llama 3.2 1B Instruct | $0.027 / $0.201 | 8 models |
| Together AI | 22 | DeepSeek V4 Flash 0731 | $0.14 / $0.28 | 4 models |
| Nebius | 17 | Qwen3-30B-A3B-Instruct-2507 | $0.10 / $0.30 | 3 models |
| Scaleway | 12 | GPT OSS 120B | $0.1682 / $0.6729 | 1 models |
| Fireworks AI | 9 | GLM-5.3-FlashX | $0.15 / $0.50 | 5 models |
| Baseten | 8 | DeepSeek V4 Flash 0731 | $0.13 / $0.26 | 1 models |
| Groq | 7 | Llama-3.1-8B-Instruct | $0.05 / $0.08 | 0 models |
| Cerebras | 2 | GPT OSS 120B | $0.35 / $0.75 | 0 models |