Model pricing
AI model pricing: cheapest LLMs by token cost
AI model prices span a wide range. The cheapest model with a measured intelligence score is GPT OSS 20B at $0.124 per 1M tokens blended, while the most capable frontier models run to around $22.00. This index ranks the cheapest scored models so you can find a budget option without dropping to an untested one. Output tokens cost several times more than input, so weight output for your workload, and verify the provider source before production billing.
Estimate your model cost - Enter your token mix and volume to get monthly spend →
Cheapest AI models by blended token cost
Chat-capable models that post a measured Artificial Analysis intelligence score, ranked by a 70% input / 30% output blend from the ByteCosts pricing index.
| Model | Provider | Input /1M | Output /1M | Blend /1M | AA index |
|---|---|---|---|---|---|
| GPT OSS 20B | Vertex | $0.070 | $0.250 | $0.124 | 14.9 |
| Gemini 2.0 Flash-Lite | $0.075 | $0.300 | $0.142 | 8.8 | |
| solar-mini | Upstage | $0.150 | $0.150 | $0.150 | 6.2 |
| GPT-5 Nano | OpenAI | $0.050 | $0.400 | $0.155 | 19.9 |
| Devstral Small | Mistral | $0.100 | $0.300 | $0.160 | 11.8 |
| Mistral Small 3.2 | Mistral | $0.100 | $0.300 | $0.160 | 10.6 |
| GPT OSS 120B | Vertex | $0.090 | $0.360 | $0.171 | 23.8 |
| DeepSeek V4 Flash | DeepSeek | $0.140 | $0.280 | $0.182 | 40.3 |
| Gemini 2.0 Flash | $0.100 | $0.400 | $0.190 | 12.3 | |
| Gemini 2.5 Flash-Lite | $0.100 | $0.400 | $0.190 | 15.1 | |
| GPT-4.1 nano | OpenAI | $0.100 | $0.400 | $0.190 | 9.6 |
| Mistral 7B | Mistral | $0.250 | $0.250 | $0.250 | 2.1 |
Frequently asked questions
What is the cheapest AI model?
The cheapest model with a measured intelligence score is GPT OSS 20B at $0.124 per 1M tokens blended (70% input / 30% output). Untested, ultra-cheap models are excluded so the ranking stays useful. The cheapest model still has to clear your quality bar, so confirm it on your own task before committing.
Why is the cheapest model not always the right choice?
Output tokens cost several times more than input, and a weaker model can take more turns, retries, or tokens to finish a task, so a higher-quality model can be cheaper per completed job. Match the model to the workload, not just the headline rate.
How is this different from the comparison and provider pages?
This page ranks individual models by token cost across providers. The compare pages put two frontier flagships head-to-head on a workload, and the provider directory ranks each provider by the cheapest model it offers.
Where do these model prices come from?
Each rate comes from the provider's official pricing, normalized into the ByteCosts pricing index and dated (updated July 9, 2026). Rows carry a source link and a confidence grade. Prices are list prices, so verify the provider source before production billing decisions.
AI model pricing: cheapest LLMs by token cost. ByteCosts. Updated July 9, 2026. https://bytecosts.com/pricing/