LiveBrowseComp - Avg@4 - Movies: leaderboard
Evaluates deep web research on the LiveBrowseComp paper snapshot using recent long-tail facts and a common search, page retrieval and Python scaffold. Scores are accuracy averaged over four attempts (avg@4), not pass@4. The shared budget is 250 steps and 256k context; source model labels are retained. Domain results are slices of the same paper evaluation. This series reports LiveBrowseComp - Avg@4 - Movies using Avg@4 Accuracy (%).
Metric: Avg@4 Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 48 | #83 |
| 2 | Claude Sonnet 4.6 | 45 | #94 |
| 3 | Gemini 3.1 Pro (Preview) | 41.9 | #64 |
| 4 | DeepSeek V4 Pro | 38.7 | #112 |
| 5 | GLM-5.1 | 34.3 | #109 |
| 6 | DeepSeek V3.2 | 32.5 | #210 |
| 7 | GLM-5 | 30.1 | #149 |
| 8 | Kimi K2.6 | 26.7 | #113 |
| 9 | Kimi K2.5 | 25 | #152 |
| 10 | MiniMax-M2.5 | 16.9 | #316 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=livebrowsecomp-avg-4-movies · How It Works · Data refreshed daily, snapshot 2026-11-05.