Muse Glimmer Lags in Agentic Evals, but Leads in Tool Use and Hallucination Control

ArtificialAnlys · x · 2026-08-11

According to Artificial Analysis, Muse Glimmer's gaps against its class primarily concentrate in agentic evaluations. It scores 953 Elo on GDPval-AA v2, trailing Qwen3.6 27B (1141), Gemini 3.5 Flash-Lite (1141), and Kimi K2.5 (1004). It also falls behind on Terminal-Bench v2.1 (52% vs. Qwen's 61%).

However, the model excels in hallucination control, scoring 82% on AA-Omniscience compared to 49% for Qwen3.6 27B and 34% for Flash-Lite (lower is better). Furthermore, in agentic tool use, Muse Glimmer scores strongly on Tau3-Banking at 24%, significantly beating Gemini 3.5 Flash-Lite (18%) and Qwen3.6 27B (17%). It outperforms its size-twin Gemma 4 31B across all these measures.

Original post →

More from Models

Models channel →