We may earn a commission if you buy through links on this page. It never changes a score or a ranking. How we make money
Pricing is quote-based.
DeepInfra lists no free plan or free trial. Its Pay-as-you-go inference option charges by model usage, without long-term contracts or upfront costs. Some language models have separate input and output token rates; most other models are billed by inference execution time. The cheapest listed token-priced example is $0.02 per 1 million input tokens and $0.04 per 1 million output tokens. Check the live model rate before forecasting a workload.
01Every plan, line by line
| Plan | Billed annually | Billed monthly | Built for |
|---|---|---|---|
| Pay-as-you-go inference | — | Quote | Teams needing API-based model inference |
02What the plan grid doesn't tell you
Different model rates
The catalog does not have one flat rate: compare the selected model’s current price before routing traffic or forecasting monthly usage.
Input and output charges
Some models price input and output tokens separately. A workload with long outputs can cost more than input-only estimates suggest.
Execution-time billing
Most models use inference-time billing rather than token rates. Confirm how the selected model measures billable execution.
Model-specific cached rates
Some model listings show a separate cached-token price. Check whether your request pattern qualifies and how cached usage is billed.
03Which plan should you buy?
| If you are… | Get | Why |
|---|---|---|
| Side project or prototype | Pay-as-you-go inference | No long-term contract or upfront cost is listed; account for the lack of a free tier. |
| Startup in production | Pay-as-you-go inference | Model-based billing can fit variable usage, but estimate costs with the specific models and billing units you plan to use. |
| Team at scale | Pay-as-you-go inference | There is no listed volume plan. Ask the vendor about expected throughput, model rates and any commercial terms for your workload. |
04Is DeepInfra worth the money?
DeepInfra’s pay-as-you-go model is worth considering when you want hosted access to a broad model catalog and can track usage by model. There is no listed free tier, so even development traffic is subject to the selected model’s charges. Token-priced examples start at $0.02 per million input tokens and $0.04 per million output tokens, while most other models use execution-time billing. Self-hosting is not listed as an option. Compare actual model rates and ask about limits before committing a production workload.
DeepInfra pricing FAQ
What is the cheapest listed DeepInfra price?
The cheapest listed token-priced example is $0.02 per 1 million input tokens and $0.04 per 1 million output tokens. It is an example rate, not a platform-wide starting price; confirm the current price for the model you intend to use.
Does DeepInfra offer a free tier?
No free plan or free trial is listed. DeepInfra’s Pay-as-you-go inference option bills by model usage. For testing, choose a model and check its current rate, then set your own usage limits and alerts.
Are all DeepInfra models priced per token?
No. The pricing information says some language models are billed per input and output token, while most other models are billed by inference execution time. Check the selected model’s pricing unit before projecting costs.
Does DeepInfra charge a monthly subscription?
The listed pricing is pay-as-you-go, with no monthly or annual billing shown. DeepInfra says it has no long-term contracts or upfront costs. Confirm any separate commercial terms directly with the vendor if you are negotiating a larger deployment.