AI BriefingDeepSeek-V3-0324 Model Released
Key point
DeepSeek has released DeepSeek-V3-0324, a model with improved performance and licensing.
Details
DeepSeek-V3-0324, an updated version of DeepSeek-V3, has been released. The core of this update is that, along with performance improvements, the license has been changed from the existing custom license to the MIT License, increasing its commercial usability.
The key benchmark performance changes are as follows:
- MMLU-Pro: 75.9 → 81.2 (+5.3)
- GPQA: 59.1 → 68.4 (+9.3)
- AIME (Math): 39.6 → 59.4 (+19.8)
- LiveCodeBench (Coding): 39.2 → 49.2 (+10.0)
This model is on par with GPT-4.5, and shows overall stronger performance than Claude-3.7-Sonnet. In particular, frontend web development, Function Calling accuracy, and Chinese writing and search capabilities have significantly improved.
The model can currently be tested immediately via Hugging Face's Inference Providers (Fireworks, Hyperbolic, Novita), and is also supported in TGI and SGLang environments.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Step 5 Preview Released: 600B Parameter MoE Model Highlights Agentic Tasks and Financial Analysis PerformanceInception Releases Ultra-Fast LLM 'Mercury 2.5' Generating 770 Tokens Per Second