AlphaFold3 Is No Longer the Top Model
anshulkundaje · x · 2026-07-15
The author argues that AlphaFold3 is no longer the best model.\n\nOver the past year, numerous companies have claimed their new models match or surpass AF3 on crucial structure prediction tasks. However, under independent verification, many failed to achieve the benchmark numbers they initially claimed, leading the author to view recent results from several teams with caution.\n\nThe post specifically mentions being initially skeptical of results published by teams behind ESMFold2, Protenix-v2, and OpenDDE, as they released benchmarks showing superiority over AF3 within a matter of weeks of each other.
More from Models
- Local benchmark: Flint matches Qwen3.6 quality at 50/52 while running 3.6x faster on M4 Pro — RichardSocher · 2026-10-27
- Mirendil debuts Fern, a from-scratch model with up to 10M token context built for agentic AI R&D — shayan_ · 2026-10-27
- ICML 2026 position paper: LLM 'thinking traces' are a harmful anthropomorphism — rao2z · 2026-10-27
- One stray 'u' and Claude starts spelling like a Brit — cantrell · 2026-10-27
- Cohere's Invent-A-Dataset paper: zero-seed synthetic data beats frontier APIs on quality and diversity — idanbeck · 2026-10-27
- ARC Prize redesigns leaderboard: highest reasoning effort by default, CSV export — mhmazur · 2026-10-27