Compute
Hosted open-model inference
Pay-per-token APIs for open-weight models.
What it is
Inference providers host open-weight models (Llama, Qwen, DeepSeek, Mistral, gpt-oss…) behind an API. The same model is often available from many hosts at different prices, speeds and quantisation levels.
How pricing works
Per token like LLM APIs. Ranking uses output price for the specific model on that host.
We compare everything in one unit: USD per 1M output tokens.
Typical uses
- Cheaper alternatives to proprietary LLMs
- Avoiding vendor lock-in
- High-throughput batch generation
What to compare
- Tokens per second
- Quantisation (FP8/INT4 lowers quality slightly)
- Rate limits
- Context length offered
Watch out for
- Some hosts serve heavily quantised versions; quality may differ from the reference model.
All 15 APIs, cheapest first
Free to browseTick 2 to 4 APIs to compare them side by side.
| API | $ / 1M out tokens | Model | Output speed | Precision | Quality | Price data | Compare |
|---|---|---|---|---|---|---|---|
| DeepSeek-V4-Flash (DeepInfra) DeepInfra | $0.18 in $0.09 | deepseek | — | — | – | Verified Oct 6, 2026 | |
| Llama 3.3 70B (Novita) Novita AI | $0.40 in $0.135 | llama | — | — | – | Verified Oct 6, 2026 | |
| Llama 3.3 70B (Hyperbolic) Hyperbolic | $0.40 in $0.40 | llama | — | fp8 | – | Unverified sample | |
| gpt-oss-120B (Cerebras) Cerebras | $0.69 in $0.25 | gpt-oss | 2,500 tok/s | — | – | Unverified sample | |
| Llama 3.3 70B (Groq) Groq | $0.79 in $0.59 | llama | 300 tok/s | fp8 | 74 | Unverified sample | |
| Llama 3.3 70B (Together) Together AI | $0.88 in $0.88 | llama | 120 tok/s | fp8 | 74 | Unverified sample | |
| Qwen (Together) Together AI | $1.20 in $0.60 | qwen | 90 tok/s | fp8 | 80 | Unverified sample | |
| DeepSeek V4.1 Flash (Novita) Novita AI | $1.20 in $0.30 | deepseek | — | — | – | Verified Oct 6, 2026 | |
| DeepSeek V4.1 Flash (Baseten) Baseten | $1.20 in $0.30 | deepseek | — | — | – | Verified Oct 6, 2026 | |
| DeepSeek (Fireworks) Fireworks AI | $2.19 in $0.56 | deepseek | 80 tok/s | fp8 | 82 | Unverified sample | |
| Llama 3.1 70B (Workers AI) Cloudflare Workers AI | $2.25 in $0.293 | llama | — | — | – | Unverified sample | |
| MiniMax-M3 (SambaNova) SambaNova | $2.40 in $0.60 | — | — | — | – | Verified Oct 6, 2026 | |
| Kimi K2.5 (Novita) Novita AI | $3.00 in $0.60 | kimi | — | — | – | Verified Oct 6, 2026 | |
| DeepSeek V3.2 (SambaNova) SambaNova | $4.50 in $3.00 | deepseek | — | — | – | Verified Oct 6, 2026 | |
| GLM-5.3 Fast (Baseten) Baseten | $6.60 in $2.10 | glm | — | — | – | Verified Oct 6, 2026 |