Nex-N2.6 Release - AutomationBench: leaderboard

Nex-N2.6 Release - AutomationBench — Nex-N2.6 official release card,model-level results,not a separate evaluation of this FP8 export. OSWorld/TerminalBench2.1/AutomationBench use source reported scores with unreported exact harness/effort/snapshot. Internal reliability uses known-issue checkpoint replay fixing prior conversation/tool outputs; tool adherence checks declared tool lists. Exact denominators and harness versions are unreported. No substitution of vendor recommended inference defaults as measured settings. Nex fine-tunes are distinct from their GLM base; earlier Nex2.5Pro stays separate. Metric: Source reported score (0–100).

Metric: Source reported score (0–100). Source: huggingface.co. Saturation forecast: Around May 2027. 4 models tracked. Updated 15 Nov 2026.

Top models

#ModelScoreOverall rank
1DeepSeek V4.1 Flash54.8#62
2GLM-5.3 Flash48.8#125
3Nex-N2.645.7#48
4Nex-N2.5-Pro44.2#246

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=nex-n2-6-release-automationbench · How It Works · Data refreshed daily, snapshot 2026-11-16.