Nate@nate_512

Seguir

Offloop multi-agent beats Claude Code and Codex

A team of four just outperformed Claude Code and Codex on advanced reasoning tasks. Offloop launched a multi-agent harness today that is achieving impressive results on the toughest workflow benchmarks in the field. They achieved state-of-the-art performance on: → GDPval (producing docs, sheets, and PDFs) → GDP.pdf → JobBench And they're doing it at 5x lower cost.

1mo

Nate@nate_512

Seguir

Offloop multi-agent beats Claude Code and Codex

A team of four just outperformed Claude Code and Codex on advanced reasoning tasks. Offloop launched a multi-agent harness today that is achieving impressive results on the toughest workflow benchmarks in the field. They achieved state-of-the-art performance on: → GDPval (producing docs, sheets, and PDFs) → GDP.pdf → JobBench And they're doing it at 5x lower cost.

1mo