Tencent Hy4 benchmark charts: 78.9 on OSWorld, closing on Claude Opus 5.5
heyshrutimishra · x · 2026-11-04
Tencent's Hy4 scores 78.9 on OSWorld 2.1, within a few points of Claude Opus 5.5 (81.8), beating GLM 5.3, DeepSeek-V4.1-Flash, Kimi K3.1 Pro and GPT 6.1 Sol. The first Hy model with vision that can drive a computer, it will reportedly ship with open weights. Shared charts cover 12 agent/coding benchmarks where Hy4 leads the open field.
More from coding & agent
- Replit's User Ops lead built an AI app to prioritize and route support tickets — amasad · 2026-11-04
- Autonomous agent designs target-specific MSAs to improve AlphaFold 3 complex prediction — SJTU · 2026-11-04
- NVIDIA paper: frequent context compaction preserves accuracy, long-context models may be unnecessary — nvidia · 2026-11-04
- Vibe-coded with Opus 5.5 and Meshy: five stained-glass 3D scenes with switchable lighting — Pristine_Good7326 · 2026-11-04
- Why LLM-Written Docs Fail: Models Optimize for Punchiness, Not Readability — mervenoyann · 2026-11-04
- OpenAI DevDay demo shows forking agent task history at any step — paw_lean · 2026-11-04