Today's frontier
Claude Fable 5.1 and GPT-6 Astra share the frontier lead
Anthropic and OpenAI are tied on the Artificial Analysis intelligence index.
ModelIntelligence (0–100)
Practical starting points from real OpenRouter usage, production performance, and independent benchmarks.
Practical starting points from real OpenRouter usage, production performance, and independent benchmarks.
Anthropic and OpenAI are tied on the Artificial Analysis intelligence index.
The most capability per dollar of list price.
| Rank | Model | Coding pctl | Agentic pctl | Tokens | Intel pctl |
|---|---|---|---|---|---|
| 1 | 55 | 49 | 90B | – | |
| 2 | 60 | – | 1.4T | – | |
| 3 | 79 | 78 | 12.4T | 69 | |
| 4 | 84 | 94 | 12.2T | 85 | |
| 5 | 64 | 62 | 5T | 50 |
Highest median output speed across multiple providers.
| Rank | Model | Throughput | TTFT p50 | Providers |
|---|---|---|---|---|
| 1 | 746 tok/s | 233ms | 22 | |
| 2 | 396 tok/s | 184ms | 16 | |
| 3 | 352 tok/s | 114ms | 13 | |
| 4 | 231 tok/s | 555ms | 10 | |
| 5 | 227 tok/s | 1.1s | 7 |
Pin one model ID and always get the newest release in that family.
| Rank | Model | Family tokens | Context |
|---|---|---|---|
| 1 | 19.4T | 1.1M | |
| 2 | 6.6T | 1M | |
| 3 | 516.7B | 1M | |
| 4 | 12.2T | 1.3M | |
| 5 | 18T | 1.3M |
The most-used models with a free endpoint.
| Rank | Model | Tokens/wk | Context |
|---|---|---|---|
| 1 | 3.7T | 1M | |
| 2 | 1.2T | 262K | |
| 3 | 1T | 1M | |
| 4 | 851.4B | 262K | |
| 5 | 332.7B | 262K |
Generate videos from text and image prompts across the leading model providers.
Start with a focused set of models for the job you need to do.