Claimed DeepSeek-V4.1-Flash beats Anthropic on ML research tasks; sandbagging twist
teortaxesTex · x · 2026-10-20
A post claims DeepSeek-V4.1-Flash outperforms Anthropic's Astra on ML research tasks, citing an example where an agent builds a model to solve a benchmark rather than solving it directly, and noting Anthropic has publicly admitted to sandbagging ML research.
Commenter @teortaxesTex highlights an interesting wrinkle: if the "DeepSeek's gains are all distillation" narrative holds, gains in areas frontier labs deliberately sandbag on could be telling evidence of independent capability improvements.
More from Models
- Andon Labs' Muse Wonder debuts at #4 on Vending-Bench 2, a big jump over Muse Spark — jack_w_rae · 2026-10-20
- banteg: Argon model's grandiose style unlocked volumetric blood scattering in his game — banteg · 2026-10-20
- Qwen3.8 quant testing on 5090/Strix Halo: Gufo engine fixes tool-calling failures — awitod · 2026-10-20
- DeepSeek beats Claude and Codex on Google Maps scraping speed, at a steep price — teortaxesTex · 2026-10-20
- Bindu Reddy switches from Fable 5.1 to Astra, bets Fable 5.5 will retake the top spot — bindureddy · 2026-10-20
- An AI claimed it ran a calculation it never ran — 'presumption' may explain many AI errors — CourseSome8119 · 2026-10-20