Skip to content

Benchmark

WideSearch · Atria paper score

Scores are percentages. Higher is better.

Models measured
4
Scores recorded
4
Sources
1
IQ weight
No weight published

Score distribution

81.9%
Mean
82.7%
Median
1.4%
Std dev
79.6%
10th percentile
83.3%
90th percentile
79.6%
Min
79.6%83.3%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1GPT-5.6 SolAtria Dawn Preview official card / report Table 1 · published comparison run · model snapshots and reasoning tiers not specified unless stated · in-house search/visit/python harness · 500 steps · 256K context / 64K max output · score subtype unstated· Reported by Shanghai AI Laboratory· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.3%
2GLM 5.3Atria Dawn Preview official card / report Table 1 · published comparison run · model snapshots and reasoning tiers not specified unless stated · in-house search/visit/python harness · 500 steps · 256K context / 64K max output · score subtype unstated· Reported by Shanghai AI Laboratory· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82.7%
3Qwen3.8 MaxAtria Dawn Preview official card / report Table 1 · published comparison run · model snapshots and reasoning tiers not specified unless stated · in-house search/visit/python harness · 500 steps · 256K context / 64K max output · score subtype unstated· Reported by Shanghai AI Laboratory· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.9%
4Kimi K3Atria Dawn Preview official card / report Table 1 · published comparison run · model snapshots and reasoning tiers not specified unless stated · in-house search/visit/python harness · 500 steps · 256K context / 64K max output · score subtype unstated· Reported by Shanghai AI Laboratory· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.6%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.