Qwen3.6-27B or 35B-A3B? A Clear Guide to Choosing the Right Model
I analyze Qwen3.6-27B vs 35B-A3B specs to guide open-source adoption. I think benchmarking methodology remains opaque.
Models & Benchmarks
Releases, papers, SOTA benchmarks
Filter by category
GPT-5.2 Pro Independently Proves 45-Year Number Theory Conjecture; Terence Tao: No Errors Found
GPT-5.2 Pro independently proved a 45-year number theory conjecture, with Terence Tao confirming no errors found.
OpenAI's Top Reasoning Expert Departs After Building o3, o1, GPT-4, and Codex
OpenAI's top reasoning expert leaves after building o3/o1/GPT-4/Codex. This exodus signals deep instability in their core R&D team.
Even $30 Billion Might Not 'Recreate GPT-4'? NUS's You Yang Reveals the Truth Behind AI Growth Bottlenecks
You Yang argues $30B won't recreate GPT-4. I read his analysis on AI bottlenecks and 2025–2026 industry extensions.
Tsinghua's Sun Maosong: Big Tech Can Scale, Others Should Focus on Vertical Applications | MEET2026
I read Tsinghua's Sun Maosong at MEET2026: Big Tech scales, others target verticals. My read: Fragmented benchmarks obscure true capability gaps.
China's AI Model Lands on Nature Cover! DeepSeek Reveals R1 Training Costed Just $2 Million
I read the Nature cover story. It highlights DeepSeek R1’s $2M training cost. This challenges Western AI economics. I question if this pricing is sustainable for competitors.
2025 The Bund Summit: Ant Group and Renmin University Release First Native MoE Diffusion Language Model, Open Source Coming Soon
I read the roadmap. Lab demos dazzle; deployed units must pay bills. I follow the release.
Winning an IMO Gold Medal with Prompt Engineering: Tsinghua Alumni Uncover New Findings, Proving Academia Can Rival Tech Giants Without Heavy Spending
I see Tsinghua alumni challenging Big Tech with an IMO-grade model, proving academia can rival giants without heavy spending.
7B Model's 'Emotional Intelligence' Rivals GPT-4o: Tencent Breaks Through Open-Domain RL Challenge, Scores Jump Fivefold
Tencent claims its 7B model rivals GPT-4o in 'emotional intelligence' via RLVER. I read the release; the fivefold score jump demands rigorous reproducibility checks before accepting this breakthrough.
Huawei's CloudMatrix Paper Reveals New Paradigm for AI Data Centers, Surpassing H100 in Inference Efficiency
I read Huawei’s CloudMatrix paper. It proposes a new AI data center paradigm surpassing H100 inference efficiency. This shifts the competitive landscape for 2025-2026 infrastructure.