DeepInfra API pricing (July 2026): 4 current models range from $0.080 to $0.150 per 1M input tokens and $0.180 to $0.600 per 1M output tokens. The cheapest paid model is Qwen3 32B at $0.080/1M input; the priciest is Llama 4 Maverick at $0.150/1M input / $0.600 output. Every price links to DeepInfra's official pricing page and refreshes twice daily.
About DeepInfra
Serverless inference platform serving 100+ open-weight models (DeepSeek, Llama, Qwen, Gemma, Mistral) at pay-per-token rates — one of the cheapest per-token hosts for open models. No subscription or minimums.
All DeepInfra models
Sorted by blended cost (cheapest first). Prices per 1M tokens, July 2026 — every price links to DeepInfra's official pricing page.
| Model | Input | Output | Cached in | Blended | Relative cost | Context | Status |
|---|---|---|---|---|---|---|---|
| DeepSeek V4 Flash | $0.090 | $0.180 | — | $0.135 |
1M | Current | |
| Qwen3 32B | $0.080 | $0.280 | — | $0.180 |
128K | Current | |
| Llama 4 Scout | $0.100 | $0.300 | — | $0.200 |
10M | Current | |
| Llama 4 Maverick | $0.150 | $0.600 | — | $0.375 |
1M | Current |
Self-host DeepInfra's open-weight models
These models ship with open weights, so you can serve them yourself on rented GPUs instead of paying per-token API prices — often cheaper at scale.
Partner links — we may earn a commission if you sign up, at no cost to you. We only list platforms we'd recommend regardless.
Quick stats
Open-weight models
3 open-weight models available from this provider.