DataSense-Bench - TBLite Data Selection: leaderboard
Metric: Selected-data TBLite average reward. Source: datasense-bench.github.io. Saturation forecast: Not forecast. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro | 0.13 | #96 |
| 2 | Claude Fable 5.1 | 0.13 | #8 |
| 3 | Gemini 3.5 Flash | 0.12 | #56 |
| 4 | GPT-6 Astra | 0.12 | #6 |
| 5 | Kimi K3 | 0.12 | #25 |
| 6 | GPT-5.5 | 0.12 | #27 |
| 7 | Claude Opus 5 | 0.12 | #11 |
| 8 | Claude Sonnet 5 | 0.12 | #55 |
| 9 | GPT-5.6 Sol | 0.11 | #16 |
| 10 | Claude Fable 5 | 0.11 | #13 |
No result here: #3 Claude Opus 5.5, #9 GPT-6.1 Sol, #10 Claude Sonnet 5.5.
Interactive version: theaggregate.ai/benchmark?slug=datasense-bench-tblite-data-selection · How It Works · Data refreshed daily, snapshot 2026-10-12.