Tools
Compare models
Add up to four models via ?models=slug-a,slug-b — or pick from any model page.
| Spec | DeepSeek | DeepSeek |
|---|---|---|
| Params | 291B / 37B | 1.6T / 37B |
| Architecture | moe | moe |
| Context | 1M | 1M |
| Weights | ||
| License | MIT License | MIT License |
| 183.4 GBest | 1008 GBest | |
| Cheapest hosted (out/1M) | $0.28 · Novita AI | $0.87 · OpenRouter |
| Released | Apr 22, 2026 | Apr 22, 2026 |
| LiveBench | 65.5 | 71.6 |
| LiveBench Agentic Coding | 37.6 | 42.6 |
| LiveBench Coding | 69.2 | 70 |
| LiveBench Data Analysis | 68 | 74.5 |
| LiveBench Instruction Following | 63.1 | 62.4 |
| LiveBench Language | 70.1 | 78.1 |
| LiveBench Mathematics | 79.7 | 90.7 |
| LiveBench Reasoning | 70.6 | 82.7 |
| Arena Agent (IPS) | −0.03 | −0.008 |
| Arena Coding | 1483.5 | 1502.2 |
| Arena Creative Writing | 1407.4 | 1446.1 |
| Arena Hard Prompts | 1459.3 | 1480.9 |
| Arena Instruction Following | 1428.3 | 1454.7 |
| Arena Maths | 1426.6 | 1445.3 |
| Arena Text (overall) | 1435.6 | 1457.9 |
| Arena Code (WebDev) | 1576.5via High | 1446 |
