MESH - Allocentric Overall: leaderboard

MESH tests spatial hearing across synchronized receivers and moving speakers. Direction, distance, position and transcription tasks remain separate. Word error rate can exceed 100%. Overall gives equal weight to localization and per-question transcription quality; it is not an average of the printed task metrics. This board reports equal-objective question-level Overall (0-100). Higher is better.

Metric: Equal-objective question-level Overall (0-100). Source: arxiv.org. Saturation forecast: Around 2028. 10 models tracked. Updated 14 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1MESH-agentEvaluation run: V2; Qwen3-14B planner and source specialist tools31.01
2Qwen 3.5 Omni Plus (Harness: MESH FOA-sep)21.86
3Qwen 3.5 Omni Plus (Harness: MESH FOA-W)19.45
4Qwen3.8 Omni Flash15.9#115
5MiDasheng-Spatial15.36
6Qwen3 Omni 30B A3B Instruct (Harness: MESH FOA-sep)14.22
7Qwen3 Omni 30B A3B Instruct (Harness: MESH FOA-W)13.77
8JAEGER Neural6.58
9JAEGER Classical1.21
10BAT0.04

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=mesh-allocentric-overall · How It Works · Data refreshed daily, snapshot 2026-11-16.