Provider pricing
DeepInfra AI API pricing
DeepInfra is a serverless inference provider focused on open-weight models at near-cost pricing. The catalog spans Llama 4, Qwen3, DeepSeek V4, Mistral, Nemotron, plus embedding and image generation models — all per-token with no minimum commit and an OpenAI-compatible endpoint.
Founded2022
HeadquartersPalo Alto, CA
Tracked models1
Best for
Cost-sensitive teams running open-source inference at scale without dedicated capacity.
Strengths
- Some of the lowest per-token rates on Llama, Qwen, and DeepSeek
- Per-second billing on dedicated deployments
- Embeddings, rerankers, image, and audio models on the same API
- No minimum spend, no negotiation — public pricing for everything
Tracked DeepInfra models
| Model | Input $/M | Output $/M | Context | Tier | Best for |
|---|---|---|---|---|---|
| Llama 3.3 70B (DeepInfra) | $0.230 | $0.400 | 128,000 | Mid-tier | DeepInfra-hosted Llama 3.3 70B at aggressive per-token pricing |