Tag

#Hyra-1.0 benchmarks

Browsing posts matching the tag #hyra-10-benchmarks

Jul 21, 2026 • 6 min read

Tencent's Hyra Beats Human-Optimized Benchmarks — and Outperforms Claude Code Along the Way

Tencent has released Hyra-1.0, a recursive self-improvement agent that outperformed community-optimized AI benchmarks, solved 29 open math problems, and beat Claude Code's output in a 3D modeling task — all using a surprisingly simple loop.

Abhinav Ramaswamy