http://cuil.org/story.php?title=deepseek-6 Chinese AI startup DeepSeek launches DeepSeek-V3, a large 671-billion parameter mannequin, shattering benchmarks and rivaling high proprietary methods. This smaller model approached the mathematical reasoning capabilities of GPT-4 and outperformed another Chinese model, Qwen-72B. Tue, 25 Feb 2025 02:48:09 UTC en