WorkflowArena: First Benchmark for Multi-Agent Workflow Orchestration With 145 Real-World Tasks
_akhaliq · x · 2026-11-10
Einsia AI released WorkflowArena, billed as the first benchmark for autonomous multi-agent workflow orchestration on real, long-horizon work — testing whether models that spawn their own workers and write coordinating workflows can actually orchestrate them well.
- 145 tasks across 8 professional domains, from SWE and Lean proofs to law and 3D (the graphic shows 3D scenes/short films, ops, multi-defendant legal cases, financial consolidation, protein model assessment, DFT reproduction, and Lean 4 proofs)
- 25 configurations of 9 frontier models run in Claude Code and Codex harnesses
- Every delegation is recorded: deliverables are scored alongside the workflow behind them
Project page, GitHub repo, and paper are all public; key findings follow in subsequent threads.
Related event: WorkflowArena Benchmark Tests Multi-Agent Orchestration Skills(2 posts)→
More from coding & agent
- Zero-dependency MCP hub fronts nine research repos as 14 tools without merging dependency trees — Emergency-Tutor-7 · 2026-11-11
- Video Decompiler: Open-Source Skill Has Fable 5.2 Code Every Frame of a Video — brandon_galang · 2026-11-11
- Dev on Fable 5.2 coding agent going sideways: it's an alignment problem — gandamu_ml · 2026-11-11
- Ponytail, the Claude Code Plugin With 188k Stars, Makes Agents Delete Before They Write — nevrekaraishwa2 · 2026-11-11
- Ponytail adds repo-level audit that ranks issues instead of just listing them — nevrekaraishwa2 · 2026-11-11
- 53% Less Code: Ponytail Forces Agents to Ask 'Does This Need to Exist?' Before Writing — nevrekaraishwa2 · 2026-11-11