LLM API pricing for open-source models
An LLM API lets you call a hosted model like DeepSeek, Qwen or Llama per token instead of running it on your own GPU.
If our calculator says a model won't fit on your card, you can call it through a hosted API instead of buying or renting a GPU. Prices below were read from each provider's own pricing page on Oct 30 2026 (checked Oct 30 2026). Prices change, so confirm on the provider's page.
We may earn from links marked Affiliate partner (disclosure).
Providers compared
| Provider | Pricing notes (checked Oct 30 2026) | Link |
|---|---|---|
| Novita AI Affiliate partner | Pay per token. Examples: Qwen3.8 27B $0.42 in / $3 out per 1M tokens; GPT OSS 120B $0.05 / $0.25; Llama 3.1 8B Instruct $0.02 / $0.05. | LLM APIs · Pricing |
| Together AI Official site | Priced per 1M tokens, e.g. Kimi K3 $3.00 in / $15.00 out. Batch API pricing also listed. | Pricing |
| Fireworks AI Official site | Serverless inference is pay per token with prepaid credits; usage is deducted from your balance. | Pricing |
| DeepInfra Official site | Per 1M tokens, e.g. DeepSeek-V4-Flash-0731 $0.06 in / $0.18 out ($0.015 cached). | Pricing |
| OpenRouter Official site | One API for many providers; each model lists its own price per 1M input and output tokens. | Models & prices |
Prices are per 1M tokens, copied from each provider's pricing page on the date shown.
How to find the cheapest LLM API for your model
Per-token prices vary a lot by model, so start from the model you want, then compare its input and output price on each provider's page. Output tokens usually cost more than input.
API vs renting a GPU vs running locally
If the model fits your card, running it locally costs nothing per token: check any model page. For heavy, steady use, renting a whole GPU by the hour can work out cheaper: see cloud GPU pricing. For occasional use, an API is usually simplest.
Models on this site with a Novita API
- Qwen3.8 27B API Affiliate partner · our VRAM page
- DeepSeek V4.1 Flash 1017 API Affiliate partner · our VRAM page
- GPT OSS 120B API Affiliate partner · our VRAM page
- Gemma 4 31B API Affiliate partner · our VRAM page
- Gemma 4 26B A4B API Affiliate partner · our VRAM page
- Llama 3.1 8B Instruct API Affiliate partner · our VRAM page
- DeepSeek R1 0528 Qwen3 8B API Affiliate partner · our VRAM page
Frequently asked questions
API or rent a GPU?
Pick an API when you need a model occasionally or it's far too big for your card. Rent a GPU when you'll keep it busy for hours. Compare GPU rental prices and each model's VRAM on its model page.
Building agents that need to run code or a browser? See AI agent sandboxes.
Last checked: