← Hosted open-model inference
Compare 4 hosted open-model inference APIs
| Qwen (Together) Together AI | DeepSeek-V4-Flash (DeepInfra) DeepInfra | Llama 3.3 70B (Novita) Novita AI | Llama 3.3 70B (Hyperbolic) Hyperbolic | |
|---|---|---|---|---|
| $ / 1M out tokens | $1.20 Unverified sample | $0.18 Verified Oct 6, 2026 | $0.40 Verified Oct 6, 2026 | $0.40 Unverified sample |
| Quality | 80 | – | – | – |
| Speed | 70 | – | – | – |
| Model | qwen | deepseek | llama | llama |
| Output speed | 90 tok/s | — | — | — |
| Precision | fp8 | — | — | fp8 |
| Context window | 256K tokens | — | — | — |