Every model in the pool, priced at list.
Twelve families across eleven providers today, with new models routable the day they ship. Pin any id, or let a policy choose.
| Model | Context | Input | Output | Latency | Quality |
|---|---|---|---|---|---|
| Claude Fable 5.1 anthropic/claude-fable-5-1 | 1M | $10.00 | $50.00 | 1,100 ms | 97 |
| GPT-5.6 openai/gpt-5.6 | 1M | $5.00 | $20.00 | 760 ms | 95 |
| Claude Sonnet 5 anthropic/claude-sonnet-5 | 500K | $3.00 | $15.00 | 540 ms | 92 |
| Gemini 3.1 Pro google/gemini-3.1-pro | 2M | $2.50 | $10.00 | 880 ms | 90 |
| Grok 4.5 xai/grok-4.5 | 256K | $3.00 | $15.00 | 690 ms | 88 |
| DeepSeek V4 Pro deepseek/deepseek-v4-pro | 256K | $1.10 | $2.20 | 620 ms | 86 |
| Qwen 3.8 Max qwen/qwen3.8-max | 256K | $1.60 | $6.40 | 710 ms | 85 |
| Kimi K3 moonshot/kimi-k3 | 256K | $0.90 | $2.70 | 830 ms | 84 |
| Mistral Large 3 mistral/mistral-large-3 | 128K | $2.00 | $6.00 | 480 ms | 83 |
| Gemini 3.1 Flash google/gemini-3.1-flash | 1M | $0.35 | $1.40 | 340 ms | 78 |
| Llama 4 Maverick meta/llama-4-maverick | 512K | $0.60 | $1.80 | 210 ms | 74 |
| GPT-5.6 mini openai/gpt-5.6-mini | 400K | $0.40 | $1.60 | 290 ms | 76 |
Prices are per million tokens at list rate, per provider. Latency is median time to first token over the last hour. Quality is the pool score on our benchmark traffic.
How a request finds its model.
Three steps, a few milliseconds, and a line in the log for each.
-
1
Score
Every model in the pool is scored on your last ten thousand requests, per task type, by a judge model you pick.
-
2
Rank
Your policy orders the pool — cheapest above a quality floor, fastest, best, or longest context — per request.
-
3
Route
The request goes to the top model. If it degrades, the next one takes it. You see which, and why, in the log.
Try a routed request against the whole catalog.
Free for a million routed tokens a month. No card, no markup on tokens.