Grok 4.5 is Stronger but Hallucinates More
ArtificialAnlys · x · 2026-07-09
Grok 4.5 boosted its AA-Omniscience score by 8 points to 26, with accuracy jumping from 35% to 52%. However, its hallucination rate also climbed from 25% to 54%, highlighting a trade-off between increased knowledge and hallucinations.
Related event: Grok 4.5 Released with Focus on Coding and Low Cost(61 posts)→
More from Models
- Mako debuts as a web-native model for end-to-end live web execution — Scobleizer · 2026-07-22
- Gemini 3.6 Flash reportedly scores below Gemini 3.5 Flash in a benchmark leak — bindureddy · 2026-07-22
- Google DeepMind launches Gemini 3.5 Flash-Lite and starts rolling it into Search — dipanjand · 2026-07-22
- Nemotron reaches gold-medal level on IMO 2026 with a 30/42 score — kuchaev · 2026-07-22
- GPT-6 is said to be near, with OpenAI betting on faster inference and custom chips — haider1 · 2026-07-22
- US Open Model Ecosystem Lags Behind China, but Competition Drives Innovation — sudoraohacker · 2026-07-22