The leading models for each task, in the order OpenRouter ranks them from the traffic going through its gateway. No score of ours: we show that ranking as it is, without re-sorting it.
Source: OpenRouter (openrouter.ai), retrieved on 8 November 2026.
The leading models for each task, in the order OpenRouter ranks them. The figure on the right is the input price, in dollars per million tokens.
One row per model found in at least two tasks. Each cell gives its rank in the task, bluer the higher it ranks. Ranks counted after removing OpenRouter's dynamic routers.
| Model | Programming | Technology | Science | Research | Marketing | SEO | Translation | Legal | Finance | Health | Roleplay | General knowledge |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| DeepSeek V4.1 Flash 1017 | 1 | 1 | 1 | 1 | 3 | 2 | 2 | 3 | 1 | 4 | 1 | 1 |
| GLM 5.3 Flash | 3 | 3 | 2 | 8 | 2 | 1 | 5 | 6 | 3 | 3 | 5 | 7 |
| DeepSeek V4.1 Flash | 2 | 2 | 5 | 4 | 4 | — | 1 | 4 | 2 | 6 | 2 | 4 |
| GPT-6 Luna | 6 | 4 | 4 | 3 | 1 | 8 | 3 | 2 | 5 | 1 | — | 5 |
| DeepSeek V4 Flash 0731 | — | — | 6 | — | 8 | — | 6 | 8 | — | 8 | 6 | 2 |
| MiniMax M3.1 Flash | 8 | 6 | 7 | 6 | — | 6 | — | — | 8 | — | — | 8 |
| GPT-5.6 Luna | — | — | — | 2 | 6 | — | 8 | 1 | 7 | 5 | — | — |
| Claude Haiku 5.5 | — | — | — | — | 5 | 5 | 7 | 7 | — | 2 | — | — |
| GPT-6.1 Sol | — | 5 | 8 | 5 | 7 | 3 | — | — | — | — | — | — |
| DeepSeek V4 Flash 0423 | — | — | — | — | — | — | 4 | — | — | — | 3 | — |
Source: OpenRouter (openrouter.ai), retrieved on 8 November 2026.
For each task, the bar runs from the cheapest to the most expensive leading model; the blue dot is the number one, its price written on the right. Logarithmic scale.
| Task | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Programming | $0.15 | $0.051 | $0.15 | $1 | $0.1 | $0.1 | $2 | $0.15 |
| Technology | $0.15 | $0.051 | $0.15 | $0.1 | $2 | $0.15 | $0.07 | $4 |
| Science | $0.15 | $0.15 | $0.1 | $0.1 | $0.051 | $0.043 | $0.15 | $2 |
| Research | $0.15 | $0.2 | $0.1 | $0.051 | $2 | $0.15 | $0.2 | $0.15 |
| Marketing | $0.1 | $0.15 | $0.15 | $0.051 | $0.1 | $0.2 | $2 | $0.043 |
| SEO | $0.15 | $0.15 | $2 | $1.75 | $0.1 | $0.15 | $0.75 | $0.1 |
| Translation | $0.051 | $0.15 | $0.1 | $0.091 | $0.15 | $0.043 | $0.1 | $0.2 |
| Legal | $0.2 | $0.1 | $0.15 | $0.051 | $0.15 | $0.15 | $0.1 | $0.043 |
| Finance | $0.15 | $0.051 | $0.15 | $0.14 | $0.1 | $4 | $0.2 | $0.15 |
| Health | $0.1 | $0.1 | $0.15 | $0.15 | $0.2 | $0.051 | $0.25 | $0.043 |
| Roleplay | $0.15 | $0.051 | $0.091 | $0.83 | $0.15 | $0.043 | $0.44 | $0.75 |
| General knowledge | $0.15 | $0.043 | $2 | $0.051 | $0.1 | $0.11 | $0.15 | $0.15 |
Source: OpenRouter (openrouter.ai), retrieved on 8 November 2026.
The same catalogue, sorted this time on the Artificial Analysis intelligence index, which OpenRouter publishes for some of the models. Compare it with the lists above: the two rankings do not always match.
The 20 highest-scoring paid models: Artificial Analysis intelligence index against input price, one variant per model; the top scorer in solid blue. Price on a logarithmic scale.
| Model | Index | Input price |
|---|---|---|
| Claude Opus 5.5 | 59.1 | $4 |
| GPT-6 Astra | 56.9 | $10 |
| GPT-6.1 Sol | 54.6 | $2 |
| Claude Sonnet 5.5 | 54.5 | $2 |
| GPT-6.1 Terra | 51.3 | $1 |
| Gemini 4 Argon | 50.4 | $2 |
| Claude Fable 5.1 | 50 | $10 |
| Kimi K3.1 Pro | 46.5 | $2 |
| GPT-6 Sol | 44.2 | $2 |
| Claude Opus 5 | 44.2 | $5 |
| Muse Wonder | 43.6 | $4 |
| Muse Spark 1.3 | 36.6 | $1.25 |
| Claude Haiku 5.5 | 36.5 | $0.1 |
| Grok 4.7 | 35.5 | $2 |
| Hy4 | 34.6 | $0.42 |
| Qwen3.8 Max (0902) | 33.7 | $2 |
| GPT-5.6 Terra | 32.3 | $2 |
| GLM 5.3 | 31.7 | $0.07 |
| Grok 4.6 | 31 | $2 |
| Gemini 3.8 Flash | 30.8 | $0.83 |
Source: OpenRouter (openrouter.ai), retrieved on 8 November 2026.
Artificial Analysis index, published by OpenRouter
How much text a model holds in view at once. It decides the size of the file you can hand it in one go: a contract, a set of accounts, a codebase.
Number of models by context window size, out of 453 models. Declared capability, not actual usage.
Source: OpenRouter (openrouter.ai), retrieved on 8 November 2026.
Context window, in tokens
The capabilities the catalogue's models declare. This is what they accept, not what is actually used.
Share of the 453 models declaring each capability; a model can tick several.
Source: OpenRouter (openrouter.ai), retrieved on 8 November 2026.
The pace at which new models join the OpenRouter catalogue.
Models by month of entry into the catalogue, over the last twelve full months. Only models still in the catalogue are counted.
| Month | Models |
|---|---|
| November 2025 | 8 |
| December 2025 | 21 |
| January 2026 | 7 |
| February 2026 | 21 |
| March 2026 | 20 |
| April 2026 | 36 |
| May 2026 | 13 |
| June 2026 | 18 |
| July 2026 | 34 |
| August 2026 | 36 |
| September 2026 | 58 |
| October 2026 | 34 |
Source: OpenRouter (openrouter.ai), retrieved on 8 November 2026.
Method
Source: OpenRouter (openrouter.ai/rankings), retrieved on 8 November 2026. Rankings licensed under CC BY 4.0.
What decides an agent is the work it is given, the data it touches and where it runs. The model itself changes in one line. Let's talk about the rest.