http://cuil.org/story.php?title=deepseek-3 Chinese AI startup DeepSeek launches DeepSeek-V3, a massive 671-billion parameter mannequin, shattering benchmarks and rivaling prime proprietary methods. This smaller model approached the mathematical reasoning capabilities of GPT-4 and outperformed another Chinese model, Qwen-72B. Mon, 24 Feb 2025 23:47:40 UTC en