AGL1K - Speech Distance Error: leaderboard
AGL1K mean great-circle distance error in km, lower is better, on the half of its crowd-sourced recordings that contain human speech, where spoken language is the dominant clue to location. A refusal counts as a 10,000 km error; Beihang University, 2026.
Metric: Mean distance error on clips with speech (km). Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 1548.29 | #135 |
| 2 | Gemini 2.5 Pro | 1634.22 | #262 |
| 3 | Gemini 2.0 Flash | 2189.3 | #702 |
| 4 | Gemini 2.0 Flash (Thinking) | 2265.47 | #702 (Gemini 2.0 Flash) |
| 5 | Gemini 2.0 Flash Lite | 2296.28 | #1032 |
| 6 | Gemini 2.5 Flash | 2689.96 | #452 |
| 7 | Gemini 2.5 Flash Lite | 3742.97 | #958 |
| 8 | Qwen2.5-Omni-7B | 4540.06 | #1797 |
| 9 | Gemma 3n E4B (IT) | 5402.73 | #2220 |
| 10 | Phi-4 Multimodal Instruct | 6311.65 | #3914 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=agl1k-speech-distance-error · How It Works · Data refreshed daily, snapshot 2026-11-12.