AutomationBench
AI benchmark results and model performance
Type: percentageAgents
Recorded results
10
Average score
31.42%
Score range
13.50% - 54.80%
Leading model
54.80% - DeepSeek V4.1 Flash
Scores over time
Individual benchmark scores plotted by date.
Model results
| Organisation | Models | Reported | Top score | Info | Self reported | Source |
|---|---|---|---|---|---|---|
| DeepSeek V4.1 Flash | 10 Sept 2026 | 54.80% | Instruct model; reasoning_effort=100; Pass@1 | Yes | Source | |
| Gemini 4 Carbon | 28 Oct 2026 | 53.90% | Google-reported launch evaluation; Zapier; source chart: https://storage.googleapis.com/gweb-uniblog-publish-prod/images/gemini_4_carbon__evals-table.original.png; methodology: deepmind.google/models/evals-methodology/gemini-4-carbon | Yes | Source | |
| Gemini 4 Argon | 13 Oct 2026 | 51.30% | Google-reported launch evaluation; source chart: https://storage.googleapis.com/gweb-uniblog-publish-prod/original_images/gemini-4-argon_table_blog.gif | Yes | Source | |
| Kimi K3 | 16 Jul 2026 | 30.80% | 600-task public subset; otherwise follows the official GitHub setup. | Yes | Source | |
| Claude Opus 5 | 24 Jul 2026 | 26% | Private held-out evaluation set; max effort | Yes | Source | |
| DeepSeek V4 Flash 0731 | 31 Jul 2026 | 25.10% | Public subset; DeepSeek-V4-Flash-0731 | Yes | Source | |
| Claude Fable 5 | 09 Jun 2026 | 17.40% | Higher of Mythos 5 / Fable 5 in Anthropic's launch table | Yes | Source | |
| Claude Mythos 5 | 09 Jun 2026 | 17.40% | Higher of Mythos 5 / Fable 5 in Anthropic's launch table | Yes | Source | |
| Claude Sonnet 5 | 30 Jun 2026 | 13.50% | Zapier leaderboard private held-out tasks; max effort | Yes | Source |
