Gemini 4 Argon posts lowest hallucination rate (15%) on AA-Omniscience benchmark
import_jmr · x · 2026-10-02
On Artificial Analysis' AA-Omniscience benchmark, Gemini 4 Argon shows the lowest hallucination rate among leading models: 15%, vs Grok 4.7 at 29%, GPT-6 Astra at 45%, and Opus 5.5 at 59%. The author argues the metric matters beyond error counts: lower hallucination suggests better metacognition — the model knows where its knowledge ends, which is what autonomy rests on.
Related event: Gemini 4 Argon Leads AA-Omniscience with 15% Hallucination Rate(2 posts)→
More from Models
- AI2: AstaBrief Fast Mode Is 3.5× Faster Than Claude-Powered Mode at Similar Quality — allen_ai · 2026-10-02
- AI2 Open-Sources AstaBrief 8B, a Model That Writes Cited Research Reports 3.5× Faster — allen_ai · 2026-10-02
- Open-Source 2B Embedding Model topk-embed-v1-small Hits Frontier Ranking on Perplexity Benchmark — antoine_chaffin · 2026-10-02
- GPT-6.1 Sol Extra High ran 20+ hours across 4 threads, still 85% quota left — MatthewBerman · 2026-10-02
- Microsoft launches new AI model claiming No. 1 spot for real-time transcription accuracy — luisdans · 2026-10-02
- Google DeepMind insider says internal model "Argon" replaced Opus as their go-to model — natanielruizg · 2026-10-02