Every AI model, scored and priced.
A benchmark score only tells you half of it. Here every model’s score sits next to what it costs to run, so you can see which one is worth the money for the job in front of you.
- models scored
- 298 models scored
- with a live API price
- 147 with a live API price
- last refreshed
- 30 Sep 2026 last refreshed
Pick by use case
Choose the job. You get the model to beat, the one with the most score per dollar, and the cheapest one that comes close to the top.
Which model is most capable overall?
240 models scored · Epoch Capabilities Index · combines dozens of benchmarks · source
How it is scoredOne benchmark per use case · blended API price · points per $
Each use case ranks on one published score, so a ranking never mixes scales. Scores come from Epoch AI’s benchmark hub, which adds new models as they ship. The Epoch Capabilities Index combines dozens of benchmarks into one number with no fixed ceiling; the others are percentages. Where a model was run at several reasoning settings, its best result counts.
Scores are as published, not re-run here. Price is OpenRouter’s cheapest listed rate, blended at the input and output split in the methodology. Points per $ is the score divided by that blended rate. The budget pick is the cheapest model within 80% of the top score (95% on the index, where the gaps are smaller).
Every model, ranked
Rank by score, by score per dollar, or by cost. Search the table, sort any column, and click a model to see its prices, speed and every score we have for it.
Epoch Capabilities Index
Composite index, no fixed ceiling · higher is better
Top 18 of 240 shown · click a bar to inspect · full list below
Leaderboard
| # | Model | In / Out | |||||
|---|---|---|---|---|---|---|---|
| 1 | OpenAI | 166.6 | 6.9 | $24.00 | — | 1.1M | $10.00 / $50.00 |
| 2 | Anthropic | 165.0 | 6.9 | $24.00 | — | 1.0M | $10.00 / $50.00 |
| 3 | Anthropic | 163.6 | 6.8 | $24.00 | — | 1.0M | $10.00 / $50.00 |
| 4 | Anthropic | 162.7 | 13.6 | $12.00 | — | 1.0M | $5.00 / $25.00 |
| 5 | OpenAI | 162.4 | 2.0 | $82.50 | — | 1.1M | $30.00 / $180.00 |
| 6 | OpenAI | 162.0 | 33.7 | $4.80 | — | 1.1M | $2.00 / $10.00 |
| 7 | OpenAI | 159.3 | 29.0 | $5.50 | — | 1.1M | $2.00 / $12.00 |
| 8 | OpenAI | 159.3 | 11.6 | $13.75 | — | 1.1M | $5.00 / $30.00 |
| 9 | OpenAI | 159.1 | 1.9 | $82.50 | — | 1.1M | $30.00 / $180.00 |
| 10 | Anthropic | 158.3 | 13.2 | $12.00 | — | 1.0M | $5.00 / $25.00 |
Out, not yet scored
Released in the last 60 days and on sale, waiting on Epoch AI benchmark results.
- Mercury 2.5 · Inception Labs · 8 Sep · $0.08/1M
- GPT-6 Luna · OpenAI · 22 Sep · $0.24/1M
- GPT-6 Sol · OpenAI · 22 Sep · $4.80/1M
- Qwen3.8 Flash · Alibaba · 26 Aug · $0.26/1M
At a glance
The top of the table on score alone, and again once price is counted.
Highest score
Epoch Capabilities Index
- GPT-6 Astra166.6
- Claude Fable 5.1165.0
- Claude Fable 5163.6
- Claude Opus 5162.7
- GPT-5.5 Pro162.4
- GPT-5.6 Sol162.0
- GPT-5.6 Terra159.3
- GPT-5.5159.3
Most score per dollar
Epoch Capabilities Index points per $1 of blended 1M-token cost · models within 90% of the top score
- GLM 5.3 Flash557.4
- DeepSeek V4 Flash 0731340.0
- GPT-5.6 Luna284.2
- DeepSeek V4.1 Flash252.0
- Inkling Small210.7
- DeepSeek V4 Pro 0813138.5
- Gemini 3 Flash Preview110.4
- GLM 5.297.3
GLM 5.3 Flash scores 91% of GPT-6 Astra at $0.27 per 1M tokens against $24.00, about 88× cheaper to run.
Ask from your editor
The TokenLens MCP server puts this data inside Claude Code, Cursor or Windsurf. Ask which model to use for a task and what it will cost, without leaving the terminal.
claude mcp add tokenlens -- \ env TOKENLENS_DATA_URL=https://tokenlens.flikq.dev/api/data \ node /path/to/tokenlens-mcp/dist/index.js > which model is best value for bug fixing?
Scores from Epoch AI (CC BY), as published. Prices from OpenRouter, blended at 35% output tokens. Methodology