DeepSeek has announced its latest model, DeepSeek V3-0324, which now ranks highest among non-reasoning AI models across multiple benchmarks. The model achieved a 66% score in the GPQA Diamond benchmark, matching Claude 3.7 Sonnet and outperforming GPT-4o and Gemini 2.0 Pro Experimental. Artificial Analysis platform hailed this achievement as "a milestone for open source," marking the first time an open weights model has led the non-reasoning category. While excelling in its class, DeepSeek V3-0324 still trails behind reasoning models like DeepSeek-R1 and OpenAI's offerings.