Cheapest LLM API
- Mistral Nemo leads
- $0.03
- models qualify
- 344
Output $/Mtok · read October 21
Cheapest output, $ per million tokens
- Mistral Nemo $0.03
- Gemma 3 4B $0.05
- S Llama 3 8B Lunaris $0.05
- Ling 3.0 Flash $0.063
- Mistral Small 3 $0.08
- Llama 3.1 8B Instruct $0.08
- gpt-oss-20b $0.09
- N Nex-N2.5-Mini $0.10
As of October 21, 2026, Mistral Nemo is the cheapest paid model, at $0.03 per million output tokens, ahead of Google Gemma 3 4B ($0.05) and Sao10K Llama 3 8B Lunaris ($0.05).
| Rank | Model | Output $/Mtok | Evidence | Coding index |
|---|---|---|---|---|
| 1 | Mistral Nemo Mistral AI | $0.03 | Not benchmarked | |
| 2 | Gemma 3 4B Google | $0.05 | Not benchmarked | |
| 3 | S Llama 3 8B Lunaris Sao10K | $0.05 | Not benchmarked | |
| 4 | Ling 3.0 Flash inclusionAI | $0.063 | Not benchmarked | |
| 5 | Llama 3.1 8B Instruct Meta Llama | $0.08 | Not benchmarked | |
| 6 | Mistral Small 3 Mistral AI | $0.08 | Not benchmarked | |
| 7 | gpt-oss-20b OpenAI | $0.09 | Not benchmarked | |
| 8 | N Nex-N2.5-Mini Nex AGI | $0.10 | Not benchmarked | |
| 9 | Ministral 3 3B 2512 Mistral AI | $0.10 | Not benchmarked | |
| 10 | Reka Edge Reka | $0.10 | Not benchmarked | |
| 11 | G MythoMax 13B Gryphe | $0.11 | Not benchmarked | |
| 12 | Granite 4.0 Micro IBM Granite | $0.112 | Not benchmarked | |
| 13 | P Laguna XS 2.1 Poolside | $0.12 | Not benchmarked | |
| 14 | Qwen3.7 Flash Qwen | $0.13 | Not benchmarked | |
| 15 | Qwen3.5-9B Qwen | $0.13 | Not benchmarked | |
| 16 | Nova Micro 1.0 Amazon | $0.14 | Not benchmarked | |
| 17 | Phi 4 Microsoft | $0.14 | Not benchmarked | |
| 18 | Schematron V2 Turbo Inference.net | $0.15 | Not benchmarked | |
| 19 | Command R7B (12-2024) Cohere | $0.15 | Not benchmarked | |
| 20 | Mercury 2.5 Inception | $0.15 | Not benchmarked |
How this list is built
Generated from the dataset on every refresh. Nobody picks the order by hand.
Every model with a paid price on OpenRouter, by output price per million tokens, cheapest first; equal output prices by input price. Free listings are left out. A low price per token is not a low price per answer: the coding index and the cost per solved task are beside it where we have them.
- OpenRouter, fetched