Loading
90 days of the enterprise, pro and team plans free, no card needed. See the plans
AI model leaderboard
Compare AI models by quality, cost, latency, coding, reasoning and real-world routing performance. Two kinds of evidence, kept apart: published benchmarks for intelligence, coding and agentic work next to the price per million tokens, refreshed three times a day - and, separately, which models and targets answered the requests routed through this deployment, counted across every workspace and never shown per account. Filter by task to see who wins coding, legal or customer-support traffic.
Live routing activity
The targets that answered the requests routed through this deployment in the last 7 days. A sample this size shows what people choose here; it is not a verdict on the models - the benchmark leaderboard below is.
Daily tokens the router sent to each target, last 7 days.
Who is behind the tokens the router routed
Ollama90%
OpenAI6%
Z.ai4%Which kinds of work reach the router
Rankings count decisions and executed calls made through this platform only; no prompt text is stored and no account is identifiable. Raw numbers: /api/v1/rankings.
Benchmark leaderboard
118 models ranked by published benchmarks - the Artificial Analysis indexes from the model listing, the Hugging Face Open LLM Leaderboard, Chatbot Arena and Hub downloads - pulled by the platform three times a day and combined into one index with its coverage shown. Models with a routing target on this deployment are marked.
| # | Model | OSR index |
|---|
Benchmarks
Independent of this deployment's traffic: the 288 catalogued models that publish Artificial Analysis indexes, ranked, with their vendor list prices. Models with a routing target here are marked.
Blended list price (3 input : 1 output tokens) per million against the intelligence index. Composite of reasoning, knowledge and instruction-following evaluations.
On the frontier: Claude Opus 5.5 (batch), Claude Sonnet 5.5 (batch), MiMo-V2.6-Pro, GLM 5.3, Claude Haiku 5.5 (batch), GLM 5.3 Flash (batch), DeepSeek V4 Flash 0731, Ling 3.0 Flash.
Photo: RDNE Stock project
| Intelligence |
|---|
| Open LLM |
|---|
| Arena |
|---|
| Downloads |
|---|
| $/1M blended |
|---|
| 1 |  | 30.3 | 16.7 | - | - | 9.8M | $0.14 |
| 2 | 24.6 | 6.7 | - | - | 9M | ||
| 3 | | 29.2 | 14.7 | - | - | 8.8M | |
| 4 |  | 40.0 | 33.7 | - | - | 7.7M | |
| 5 | | 28.4 | 13.3 | - | - | 7M | |
| 6 | Qwen2.5 7B Instructopen weights | 32.7 | - | 48.6 | - | 6.9M | $0.13 |
| 7 | | 44.7 | 41.8 | - | - | 6.5M | |
| 8 | | 26.5 | |||||
| 9 |  | 33.1 | 21.5 | - | - | 5.5M | |
| 10 |  Llama 3.1 8B Instructopen weights | 20.0 | 6.9 | 37.5 | - | 5.3M | |
| 11 | Qwen3 VL 8B Instructopen weights | 25.0 | 7.3 | - | - | 5M | $0.38 |
| 12 | Llama 3.2 1B Instructopen weights | 15.4 | 4.8 | 29.6 | - | 4.6M | |
| 13 | | 40.4 | 34.3 | - | - | 4.2M | |
| 14 | | 31.2 | 18.2 | - | - | 3.7M | |
| 15 | | 27.4 | 11.6 | - | - | 3.1M | |
| 16 | 25.0 | 7.3 | - | - | 2.8M | ||
| 17 | | 33.0 | 21.4 | - | - | 2.4M | |
| 18 | 46.4 | 44.8 | - | - | 1.9M | ||
| 19 | | 33.9 | 22.9 | - | - | 1.7M | |
| 20 | | 34.9 | 24.7 | - | - | 1.5M | |
| 21 | | 31.8 | 19.3 | - | - | 1.4M | |
| 22 | 35.1 | 25.0 | - | - | 1.4M | ||
| 23 |  | 25.9 | 8.9 | - | - | 1.2M | |
| 24 | Llama 3.2 3B Instructopen weights | 19.9 | 5.7 | 38.8 | - | 1.1M | |
| 25 | | 25.1 | 7.6 | - | - | 1.1M | |
| 26 | Qwen3 30B A3B Instruct 2507open weights | 25.1 | 7.5 | - | - | 1.1M | $0.14 |
| 27 | Gemma 3 4Bopen weights | 23.5 | 4.8 | - | - | 1.1M | $0.04 |
| 28 | | 29.3 | |||||
| 29 | | 33.9 | 22.9 | - | - | 1M | |
| 30 | 45.7 | 43.6 | - | - | 988.7K | ||
| 31 | 36.7 | 27.9 | - | - | 946.3K | ||
| 32 | DeepSeek V3 0324open weights | 26.3 | 9.7 | - | - | 911.2K | $0.50 |
| 33 |  | 33.8 | 22.8 | - | - | 827K | |
| 34 | T | 35.5 | 25.7 | - | - | 820.9K | |
| 35 | | 34.7 | 24.4 | - | - | 807.7K | |
| 36 | | 27.1 | 11.1 | - | - | 794.2K | |
| 37 | | 30.9 | 17.7 | - | - | 721.8K | |
| 38 | | 28.2 | 12.9 | - | - | 719.2K | |
| 39 | Qwen2.5 Coder 32B Instructopen weights | 26.1 | 6.7 | 51.9 | - | 642.8K | |
| 40 | | 33.0 | 21.4 | - | - | 628.9K | |
| 41 | 35.7 | 26.1 | - | - | 578.3K | ||
| 42 | Qwen3 Coder 30B A3B Instructopen weights | 26.3 | 9.6 | - | - | 576.8K | $0.12 |
| 43 | Qwen3 Coder Nextopen weights | 26.0 | 9.2 | - | - | 521.3K | $0.29 |
| 44 | | 33.8 | 22.8 | - | - | 501.3K | |
| 45 | | 40.7 | 34.8 | - | - | 414.7K | |
| 46 |  Kimi K2 0711open weights | 28.0 | 12.7 | - | - | 371K | $1.00 |
| 47 | Gemma 3 27Bopen weights | 23.6 | 4.9 | - | - | 370.5K | $0.15 |
| 48 | 26.0 | 5.9 | 52.7 | - | 370.1K | $0.09 | |
| 49 | 40.0 | 33.7 | - | - | 368.9K | ||
| 50 | 36.2 | 27.0 | - | - | 360K |
Updated 6 h ago · next refresh in 1 h · refreshed 3× a day from OpenRouter listing, Hugging Face Open LLM Leaderboard, Hugging Face Hub, Chatbot Arena leaderboard.
OSR index = 40% Intelligence + 30% Open LLM + 30% Arena over the signals a model publishes.
Composite of reasoning, knowledge and instruction-following evaluations. 288 of 288 catalogued models publish it; prices are blended 3:1 list prices per million tokens.
The catalogue prints input and output prices and what three typical monthly workloads cost on each.
Free plan, no card. Fifteen thousand decisions a month with the full trace.
| $0.15 |
| $0.25 |
| $0.37 |
| $0.09 |
| $0.24 |
| - |
| - |
| 5.9M |
| $0.04 |
| $0.32 |
| $0.06 |
| $0.07 |
| $0.07 |
| $0.31 |
| $0.26 |
| $0.13 |
| $0.69 |
| $0.41 |
| $0.74 |
| $0.11 |
| $0.35 |
| $1.76 |
| $0.10 |
| $0.12 |
| $0.21 |
| - |
| - |
| 1.1M |
| $0.15 |
| $1.05 |
| $3.16 |
| $1.55 |
| $0.44 |
| $0.64 |
| $0.11 |
| $0.31 |
| $0.71 |
| $0.10 |
| $0.74 |
| $0.88 |
| $2.15 |
| $0.53 |
| $0.32 |
| $1.78 |
| $1.34 |
Hub 
