OSS Surge China/US Embodied & Agentic Efficiency

AI Breakthroughs Digest

GLM-5.1/5V-Turbo >GPT-5.4 SWE/UI/video; Qwen3.6-35B #1 Terminal/SWE/vision; ML-Master 2.0 56% MLE-Bench science agents; HY-World 2.0 OSS robotics; DeepSeek V4/GPT-OSS hardware; Cursor3 3M WAU; Tsinghua/Stanford harness 6x perf.

Sources (8)

Updated Apr 22, 2026

AI Breakthroughs Digest

This Frankenstein AI Merges Claude Opus, GLM and Qwen—And Outperforms Top Models

Moonshot AI Releases Kimi K2.6 with Long-Horizon Coding, Agent Swarm Scaling to 300 Sub-Agents and 4,000 Coordinated Steps

The Art and Science of Harness Engineering: Redefining AI Agent Performance

@omarsar0: // Towards Ultra-Long-Horizon Agentic Science // These researchers finally got long-horizon researc...

GPT-OSS 120B vs GLM-5.1 vs DeepSeek V4

Qwen Team Open-Sources Qwen3.6-35B-A3B: A Sparse MoE Vision-Language Model with 3B Active Parameters and Agentic Coding Capabilities

@omarsar0: Cool new paper from Microsoft on Skill learning for autonomous web agents.

Chinese AI Model Tops GPT with 8-Hour Coding Runs