Researcher accuses OpenAI of sneaky perplexity disclosure designed to trap Anthropic
davidmanheim · x · 2026-11-23
David Manheim calls OpenAI's perplexity-style disclosure "sneaky and dishonest": it publicly pressures Anthropic to publish analogous numbers and then explain why they aren't comparable due to tokenizer choices, instead of OpenAI flagging that caveat upfront. He says OpenAI's track record doesn't merit good-faith benefit of the doubt. The dispute sits inside a broader debate over whether a frontier-lab pacing agreement could standardize on perplexity-style compression metrics.
More from Models
- Researcher slams OpenAI's benchmark disclosure as "sneaky and dishonest" — davidmanheim · 2026-11-23
- Thinking Machines releases Inkling 2 Small, scoring 31 on AA Intelligence Index — _Hao_Zhu · 2026-11-23
- Leaked OpenAI loss ladder: unknown capabilities await at 0.68-0.7 — tallmetommy · 2026-11-23
- Developers heap praise on Gemini 4 Neon's 'most tasteful' writing — gajesh · 2026-11-23
- METR-fit extrapolation implies 66-hour task horizons for 6.2 Astra, 128 hours for internal model — eliebakouch · 2026-11-23
- $200 Google AI Ultra subscription quietly includes a 96GB Blackwell GPU for 10 days — kieranklaassen · 2026-11-23