Anthropic's Model 4 reportedly scores 178.6 on ECI benchmark, 12 points clear
koltregaskes · x · 2026-11-06
A user reports that Anthropic's Model 4 scored 178.6 on the ECI benchmark versus 166.5 for the previous leader Fable 5.1—a notable 12-point jump on a leaderboard where top models are usually separated by only a few points. Unverified.
More from Models
- Models Learn to Prompt-Inject Graders: Apollo Disclosure Shows 2/3 Graders Fooled — MariusHobbhahn · 2026-11-06
- Critic questions METR's RSI metrics: benchmaxable and unrepresentative — gleech · 2026-11-06
- "Perplexity has the best harness by FAR": why same models answer better there — ChrisUniverse · 2026-11-06
- Screen Studio gets local Qwen3-ASR and Doubao voice for accurate Chinese subtitles — vista8 · 2026-11-06
- Grok Bot on X Premium feels like a demo: user says usage pool can't last a week — nima_owji · 2026-11-06
- GPT-6.2 'Betelgeuse' leaks surface in Codex and API as OpenAI's rumored biggest leap — TheAIGRID · 2026-11-06