--- pretty_name: WebArena-Pro NeurIPS 2026 Coding Agent Traces tags: - webarena-pro - agent-trajectories - coding-agents - neurips-2026 --- # NeurIPS 2026 coding agent traces **1,500 traces** across **five agent/model cohorts**. | Agent | Model | Traces | | --- | --- | ---: | | Claude Code | Claude Opus 5.5 | 300 | | Codex | GPT 6.1 Sol | 300 | | OpenCode | Gemini 3.1 Pro | 300 | | OpenCode | Kimi 3.1 Pro | 300 | | OpenCode | Qwen 3.8 Max Prime | 300 |