AI Rate CardThe cheapest LLM APIs right now
Updated Oct 9, 2026
The same open model is often sold by a dozen providers at very different prices. The table below is built from today's checked prices: the cheapest paid offer of each text model, by input price per 1 million tokens.
| Model | Provider | Input, $ | Output, $ |
|---|---|---|---|
| Llama 3.2 1B Instruct | deepinfra | $0.005 | $0.01 |
| GPT OSS 20B | edenai | $0.009 | $0.045 |
| DeepSeek R1 Distill Llama 8B | aihubmix | $0.01 | $0.01 |
| DeepSeek R1 Distill Qwen 1.5b | aihubmix | $0.01 | $0.01 |
| DeepSeek R1 Distill Qwen 7B | aihubmix | $0.01 | $0.01 |
| Qwen3.5 0.8B | deepinfra | $0.01 | $0.05 |
| Google Gemma 2 | helicone | $0.01 | $0.03 |
| Qwen2.5-Coder-3B-Instruct | huggingface | $0.01 | $0.03 |
| Qwen2.5-Coder 7B Instruct | huggingface | $0.01 | $0.03 |
| Qwen3 4B Instruct | huggingface | $0.01 | $0.03 |
Live data: this table is built from today's prices.
How to read the table
- Input is what you pay for the text you send (the prompt, the documents, the chat history); output is what the model writes back. In a chat the history grows with every turn, so input usually dominates the bill.
- A model's page lists every provider of that model, so you can see how far the cheapest one is from the rest.
- Free tiers are left out here: they are useful for trying a model, but their rate limits rarely fit real traffic.
What to check before you switch
- Context and output limits. A cheap offer with a short context window may not fit your documents.
- Quantization. Providers may serve a compressed version of the same model; it is cheaper and can be slightly worse.
- Region and data terms. Where the requests are processed matters if you handle personal data.
- The provider's own page. Prices change; every price on this site links to its source and shows the day it was checked.
The biggest price drops this week
| Date | Model | All similar models | Provider | Component | Old → new, $ | Change |
|---|---|---|---|---|---|---|
| 02.11.2026 12:23:54 UTC | DeepSeek: DeepSeek V4 Pro 0423 | deepseek-v4-pro | trustedrouter | input tokens | $1.85 → $0.16 | -91.4% |
| 01.11.2026 12:22:04 UTC | OpenAI: gpt-oss-120b | gpt-oss-120b | openrouter | output tokens | $0.60 → $0.06 | -90.0% |
| 04.11.2026 12:20:13 UTC | Nano Banana Pro | gemini-3-pro-image | output tokens | $120.00 → $12.00 | -90.0% | |
| 01.11.2026 12:22:04 UTC | OpenAI: gpt-oss-120b | gpt-oss-120b | openrouter | input tokens | $0.15 → $0.016 | -89.3% |
| 31.10.2026 12:20:48 UTC | Z.ai: GLM 5.3 Flash | glm-5.3-flash | openrouter | input tokens | $0.15 → $0.0163 | -89.1% |
| 29.10.2026 12:21:27 UTC | Z.ai: GLM 5.3 Flash | glm-5.3-flash | openrouter | output tokens | $0.50 → $0.0742 | -85.2% |
| 04.11.2026 12:20:13 UTC | MiniMax-M3.1-Flash | MiniMax-M3.1-Flash | cometapi | output tokens | $3.20 → $0.48 | -85.0% |
| 04.11.2026 12:20:13 UTC | MiniMax-M3.1-Flash | MiniMax-M3.1-Flash | cometapi | input tokens | $0.80 → $0.12 | -85.0% |
| 02.11.2026 12:23:54 UTC | DeepSeek: DeepSeek V4 Pro 0423 | deepseek-v4-pro | trustedrouter | output tokens | $3.69 → $0.63 | -82.9% |
| 01.11.2026 12:22:04 UTC | DeepSeek: DeepSeek V4 Pro 0813 | deepseek-v4-pro-0813 | openrouter | input tokens | $0.66 → $0.12 | -82.6% |
Live data: this table is built from today's prices.
Your own workload
Type your monthly input and output tokens into "Your cost/mo" on the Models page: the list then sorts by what you would actually pay, not by the price per token.