AA-Omniscience Eval: Sol and V4 Lead in Knowledge but Are 'Bullshit Machines'
teortaxesTex · x · 2026-07-21
A user shared interesting findings from the AA-Omniscience evaluation:
- Sol is close to Fable and categorically above K3 and V4 in knowledge.
- V4 and K3 are very close in performance.
- Grok 1.5T performs great.
- V4 is a 'bullshit machine', and so is Sol.
These results prompt a discussion on the actual effects of scaling on model hallucinations and knowledge retention.
Related event: AA-Omniscience Chart Compares Accuracy and Hallucination Rates of 28 Models(2 posts)→
More from Models
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11
- GPT-5.6 writes well but is instantly forgettable, user complains — BasedRaddka · 2026-09-11
- Opus Refuses Protein Research Codebase Over 'Safety' Concerns, Dev Considers Rolling His Own — josephdviviano · 2026-09-11
- User Hails Unconfirmed 'DeepSeek 4.1 Flash' as an Inflection Point in LLMs — himanshustwts · 2026-09-11
- Terminal Bench v4: GLM-5.3 Leads at 41.9%, Kimi-K3 Underwhelms at 12.6% — Ok_Warning2146 · 2026-09-11
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11