Tools
Compare models
Add up to four models via ?models=slug-a,slug-b — or pick from any model page.
| Spec | DeepSeek |
|---|---|
| Params | 763B |
| Architecture | dense |
| Context | 1M |
| License | |
| License terms | MIT License |
| 481.2 GBest | |
| Cheapest hosted (out/1M) | $0.40 · DekaLLM |
| Released | Sep 10, 2026 |
| LiveBench | 81.11 |
| LiveBench Agentic Coding | 77.27 |
| LiveBench Coding | 80.04 |
| LiveBench Data Analysis | 79.25 |
| LiveBench Instruction Following | 70.03 |
| LiveBench Language | 81.19 |
| LiveBench Mathematics | 93.29 |
| LiveBench Reasoning | 86.69 |
| Arena Agent | 13th of 55 |
| Arena Agent · Recoverygets back on track after a command fails | 8th of 55 |
| Arena Agent · Steerabilitydoes what it was asked, and changes course when told | 36th of 55 |
| Arena Agent · Task outcomefinishes what the session set out to do | 3rd of 55 |
| Arena Agent · Tool usereaches for the right one, and does not invent one | 31st of 55 |
| Arena Coding | 1532 |
| Arena Creative Writing | 1440 |
| Arena Hard Prompts | 1501 |
| Arena Instruction Following | 1478 |
| Arena Maths | 1509 |
| Arena Text (overall) | 1477 |
| Arena Code (WebDev) | 1621 |
