ITBench-AA: leaderboard
Artificial Analysis implementation of IBM's ITBench SRE benchmark, testing AI agents on Kubernetes incident root-cause analysis from offline incident snapshots.
Metric: Average Precision at Full Recall (imported). Source: artificialanalysis.ai. Source: benchmarklist.com. Saturation forecast: Around September 2027. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (xHigh) | 45.8 | #31 (GPT-5.5) |
| 2 | Qwen 3.7 Max | 42.5 | #75 |
| 3 | Gemini 3.5 Flash (High) | 40.3 | #60 (Gemini 3.5 Flash) |
| 4 | GLM-5.1 (Thinking) | 40.3 | #97 (GLM-5.1) |
| 5 | DeepSeek V4 Pro (Reasoning, Max Effort) | 38.3 | #102 (DeepSeek V4 Pro) |
| 6 | MiMo-V2.5-Pro | 38.2 | #158 |
| 7 | Gemma 4 31B (Reasoning) | 37.3 | #200 (Gemma 4 31B) |
| 8 | Qwen 3.5 27B (Reasoning) | 35.5 | #187 (Qwen 3.5 27B) |
| 9 | GPT-5.4 Mini (xHigh) | 35.2 | #212 (GPT-5.4 Mini) |
| 10 | GPT-5.4 (xHigh) | 34.5 | #79 (GPT-5.4) |
| 11 | Qwen 3.5 397B A17B (Reasoning) | 34.1 | #146 (Qwen 3.5 397B A17B) |
| 12 | Grok 4.3 (High) | 32.7 | #188 (Grok 4.3) |
| 13 | DeepSeek V4 Flash (Reasoning, Max Effort) | 31.5 | #137 (DeepSeek V4 Flash) |
| 14 | Kimi K2.6 | 31.2 | #104 |
| 15 | Gemini 3.1 Pro (Preview) | 30.3 | #58 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=itbench-aa · How It Works · Data refreshed daily, snapshot 2026-10-20.