DeepSeek launches V3.1 with doubled context, advanced coding, and math abilities. Featuring 685B parameters under MIT Licence ...

In a quiet yet impactful move, DeepSeek, the Hangzhou-based AI research lab, has unveiled DeepSeek V3.1, an upgraded version ...

Overview DeepSeek dominates in reasoning, planning, and budgeting, proving itself the more practical and precise choice for ...

Winner: DeepSeek wins for more detailed reasoning that better fulfills the “explain your reasoning step by step” aspect of ...

DeepSeek V3.1 launches with 128k context, 685B parameters, top coding scores, and delays its R2 model due to issues with Huawei’s Ascend chips.

Overview Open-source AI models often use up to 10x more tokens, making them more expensive than expected.DeepSeek and JetMoE ...

GPT-5 failed to wow users, but its less flashy delivery signals a positive shift for AI from at-all-costs advancement to ...

In my post on large language models (LLMs) last week, I argued that the most important question about LLMs is not the outcome of a race with China or when AI will reach human-level intelligence, but ...

The bird has never gotten much credit for being intelligent. But the reinforcement learning powering the world’s most ...

Bettors on Kalshi, a prediction market, are putting their money behind Google’s Gemini to become the top AI text model of 2025, beating OpenAI's ChatGPT and xAI’s Grok.