Offloop says its 4-person multi-agent harness beat Claude Code and Codex on GDPval
rohanpaul_ai · x · 2026-07-24
- The post argues that the real capability layer is shifting from the model to the harness around it.
- Offloop says its 4-person team built a multi-agent harness that beat Claude Code and Codex on GDPval and related workplace benchmarks.
- Reported numbers:
- GDPval: 84.9 at $1.65/task
- Claude Code (Opus 4.8): 82.4 at $14.38/task
- Codex (GPT 5.6 Sol): 83.3 at $5.20/task
- The benchmarks are framed as evaluating real knowledge work across 44 occupations and 9 industries, with shell access and web browsing, using blind comparisons against human experts.
- The quoted takeaway is that lower cost per completed task matters because it lets agents run longer, do more work per dollar, and preserve margins at scale.
Related event: Offloop's Multi-Agent Framework Tops GDPval Benchmark(10 posts)→
More from coding & agent
- Spring AI 2.1.0-RC1 adds tool-call confirmation, pausing agent loops before real-data access — therealdanvega · 2026-10-26
- Morgan Stanley and US banks block third-party AI agents over credentials — thione · 2026-10-26
- METR red-teams OpenAI's agent monitoring, finds bypasses and gaps — thione · 2026-10-26
- OpenAI: coding agents weakened safety monitoring, one lied about it — thione · 2026-10-26
- ConvFill: A 0.6B Talker Buys Time While a Large Reasoner Whispers the Real Answer — andimarafioti · 2026-10-26
- 250 Builders Hack With Codex at Berlin Community Hackathon Ahead of OpenAI DevDay — paw_lean · 2026-10-26