Tools
Compare models
Add up to four models via ?models=slug-a,slug-b — or pick from any model page.
Scroll sideways to see every model →
| Spec | Qwen | Moonshot AI |
|---|---|---|
| Params | 79.7B | 1.1T / 32B |
| Architecture | moe | moe |
| Context | 262K | 262K |
| License | ||
| License terms | Apache License 2.0 | Custom licence |
| 51.7 GBmeasured | 667.5 GBest | |
| Cheapest hosted (out/1M) | $0.80 · OpenRouter | $3.35 · OpenRouter |
| Released | Jan 30, 2026 | Jun 11, 2026 |
| LiveBench | — | 68.41 |
| LiveBench Agentic Coding | — | 45.66 |
| LiveBench Coding | — | 73.96 |
| LiveBench Data Analysis | — | 62.66 |
| LiveBench Instruction Following | — | 56.29 |
| LiveBench Language | — | 77.91 |
| LiveBench Mathematics | — | 79.6 |
| LiveBench Reasoning | — | 82.81 |
| Arena Agent | — | 25th of 55 |
| Arena Agent · Recoverygets back on track after a command fails | — | 38th of 55 |
| Arena Agent · Steerabilitydoes what it was asked, and changes course when told | — | 10th of 55 |
| Arena Agent · Task outcomefinishes what the session set out to do | — | 23rd of 55 |
| Arena Agent · Tool usereaches for the right one, and does not invent one | — | 2nd of 55 |
| Arena Code (WebDev) | — | 1473 |
