Deepnote Agent Initiative Benchmark: leaderboard
Deepnote tests when a proactive workplace agent should investigate and help, and when it should remain silent. All 12 models face the same 50 synthetic company situations at source-declared high reasoning effort, using 11 service APIs and 180 seconds of model response time (tool time excluded). Two mutually blind LLM evaluators grade against a reference solution and complete action log; the0–10mean covers all50situations. Exact judge models and API snapshots are unreported. Higher is better.
Metric: Mean blind evaluator task score (0-10). Source: deepnote.com. Saturation forecast: Around February 2027. 12 models tracked. Updated 14 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-6 Astra | High | 8.97 | #6 |
| 2 | GPT-6.1 Sol | High | 8.65 | #10 |
| 3 | Claude Fable 5.2 | High | 8.17 | #1 |
| 4 | Claude Opus 5.5 | High | 7.84 | #5 |
| 5 | Claude Fable 5.1 | High | 7.5 | #15 |
| 6 | Claude Sonnet 5.5 | High | 6.65 | #19 |
| 7 | GPT-6 Luna | High | 6.38 | #99 |
| 8 | DeepSeek V4.1 Flash | High | 6.35 | #62 |
| 9 | Mistral Large 4 | High | 6.09 | #137 |
| 10 | Claude Haiku 5.5 | High | 6.05 | #70 |
| 11 | Gemini 4 Argon | High | 5.6 | #12 |
| 12 | Grok 4.7 | High | 4.51 | #47 |
No result here: #2 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=deepnote-agent-initiative-benchmark · How It Works · Data refreshed daily, snapshot 2026-11-16.