Kaggle ITBench - FinOps: leaderboard
FinOps score from IBM's ITBench on Kaggle: the percentage score (0-100, higher is better) on its financial-operations scenarios, in which an agent works on cost-management problems. The 13-model board moves in steps of about 6.7 points, so small differences are single scenarios.
Metric: Score (%). Source: www.kaggle.com. Saturation forecast: Around November 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 235B A22B 2507 Instruct | 73.33 | #315 |
| 2 | Claude Sonnet 4.5 | 66.67 | #148 |
| 3 | Gemini 3 Pro (Preview) | 66.67 | #71 |
| 4 | GPT-5.1 | 66.67 | #145 |
| 5 | Claude Opus 4.5 | 66.67 | #87 |
| 6 | GPT-5 | 60 | #105 |
| 7 | Gemini 2.5 Pro | 53.33 | #159 |
| 8 | Gemini 2.5 Flash | 53.33 | #255 |
| 9 | O3 | 53.33 | #132 |
| 10 | Mistral Small 3.1 | 53.33 | #623 |
| 11 | Claude Haiku 4.5 | 46.67 | #287 |
| 12 | Grok 4.1 Fast (Reasoning) | 46.67 | #227 (Grok 4.1 Fast) |
| 13 | DeepSeek R1 0528 | 46.67 | #232 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=kaggle-itbench-finops · How It Works · Data refreshed daily, snapshot 2026-11-05.