Tools
Compare models
Add up to four models via ?models=slug-a,slug-b — or pick from any model page.
| Spec | DeepSeek | Moonshot AI |
|---|---|---|
| Params | 685B / 37B | 2.8T / 104B |
| Architecture | moe | moe |
| Context | 164K | 1M |
| Weights | ||
| License | MIT License | Custom licence |
| 432.1 GBest | 1752.7 GBest | |
| Cheapest hosted (out/1M) | $0.40 · Novita AI | $15.00 · Novita AI |
| Released | Dec 1, 2025 | Jun 13, 2026 |
| Arena Coding | 1469.6 | 1531.5via Max |
| Arena Creative Writing | 1400.4 | 1468.8via Max |
| Arena Hard Prompts | 1447.2 | 1505.5via Max |
| Arena Instruction Following | 1420.6 | 1479.7via Max |
| Arena Maths | 1428.5 | — |
| Arena Text (overall) | 1425 | 1485.3via Max |
| Arena Code (WebDev) | 1323.4 | 1675.6via Max |
| SWE-bench Verified | 70via mini-SWE-agent | — |
| LiveBench | — | 79.2 |
| LiveBench Agentic Coding | — | 62.2 |
| LiveBench Coding | — | 81.5 |
| LiveBench Data Analysis | — | 78.7 |
| LiveBench Instruction Following | — | 71.4 |
| LiveBench Language | — | 85.5 |
| LiveBench Mathematics | — | 84.4 |
| LiveBench Reasoning | — | 90.7 |
| Arena Agent (IPS) | — | 0.099via Max |
