TimeLitmus - Traffic Natural Prediction: leaderboard
Metric: Natural accuracy (%; 420 Traffic records of TraffiDent roadway incidents with three ordered response classes from observed post-report buckets; ten LLMs with deterministic decoding where available, invalid or unparsable outputs scored incorrect). Source: arxiv.org. Saturation forecast: Around 2034. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen Plus | 42.1 |
| 2 | DeepSeek R1 0528 | 41 |
| 3 | Qwen 3.5 9B | 40.7 |
| 4 | Qwen 3.5 4B | 38.6 |
| 5 | Claude Sonnet 4.6 | 37.9 |
| 6 | MiniMax-M3 | 37.1 |
| 7 | DeepSeek V4 Flash | 36.9 |
| 8 | Gemini 3.5 Flash | 35.2 |
| 9 | GLM-5 | 35 |
| 10 | GPT-5.4 | 34.5 |
Interactive version: theaggregate.ai/benchmark?slug=timelitmus-traffic-natural-prediction · How It Works · Data refreshed daily, snapshot 2026-09-26.