How much cheaper the same AI got
- DeepSeek V4.1 Flash 1017
- 606
- Kimi K3.1 Pro
- 561
- Mistral Large 4
- 555
Rising on Hugging Face · read November 5
- New Ling 3.0 Tiny November 5
- New R ricursiveAI/Ripple-65B-Preview November 4
- New Hy4 November 4
- New S Crouton Alpha November 3
As of November 5, 2026, the cheapest model that scores at least 141.16 on the Epoch Capabilities Index, the level of OpenAI o1, costs $0.26 per million output tokens (Qwen3.5-Flash). At today's prices, the cheapest model released by the end of Q2 2025 that reaches this level, Google Gemini 2.5 Flash, costs $2.50; the cheapest one now costs 90% less.
Cheapest model at OpenAI o1's level, quarter by quarter
| Quarter | Cheapest model at that level | Released | Output price | Change vs Q2 2025 |
|---|---|---|---|---|
| Q4 2026 | Qwen3.5-Flash | February 25, 2026 | $0.26 | -90% |
| Q3 2026 | Qwen3.5-Flash | February 25, 2026 | $0.26 | -90% |
| Q2 2026 | Qwen3.5-Flash | February 25, 2026 | $0.26 | -90% |
| Q1 2026 | Qwen3.5-Flash | February 25, 2026 | $0.26 | -90% |
| Q4 2025 | DeepSeek V3.2 Exp | September 29, 2025 | $0.41 | -84% |
| Q3 2025 | DeepSeek V3.2 Exp | September 29, 2025 | $0.41 | -84% |
| Q2 2025 | Google Gemini 2.5 Flash | June 17, 2025 | $2.50 | Baseline |
Prices are today's OpenRouter list prices of each model, not the price on its launch day, which we only record from October 2026. The index is Epoch AI's own; models it has not scored are left out.
Current prices by release year
| Release year | Priced models | Minimum output price | Median output price | Maximum output price |
|---|---|---|---|---|
| 2026 | 198 | Free | $2.04 | $180.00 |
| 2025 | 110 | $0.075 | $1.25 | $168.00 |
| 2024 | 31 | $0.05 | $0.72 | $15.00 |
| 2023 | 5 | $0.355 | $0.75 | $60.00 |
Latest dataset: November 5, 2026. USD per million output tokens. Includes listed models with a known release date and price, including free models. Retired models and missing prices are excluded. Release year is not the date the price was recorded.