Alibaba's 35-Hour AI Agent Test #shorts

Firethering reported on May 25, 2026 that Alibaba's Qwen3.7-Max ran a 35-hour autonomous kernel optimization task on unfamiliar T-Head ZW-M890 PPUs, making 1,158 tool calls and 432 kernel evaluations. The reported 10x speedup and benchmark table come from Alibaba's own evaluation via Qwen Blog, so independent reproduction is still needed. Self-reported Terminal Bench 2.0: Qwen3.7-Max 69.7, Claude Opus 4.6 65.4, DeepSeek V4 Pro 67.9. Source: https://firethering.com/alibaba-qwen3-7-max-autonomous-agent/

Видео Alibaba's 35-Hour AI Agent Test #shorts канала Census Drift