Qwen 3.7 Max hits 10x kernel speedup

The "agentic marathon" is the new benchmark, shifting focus from chat accuracy to sustained autonomous reasoning and tool use.

• –Qwen 3.7 Max made 1,158 tool calls and 432 evaluations over 35 hours without human intervention.
• –Rewrote the SGLang Triton attention kernel to achieve a 10.0x geometric mean speedup on the new Zhenwu M890 processor.
• –Features a 1M-token context window and full Model Context Protocol (MCP) integration for complex workflows.
• –Performance in GPQA Diamond (92.4) surpasses competitors like GLM 5.1 and DeepSeek V4.
• –Proprietary release via API ($2.50/$7.50) signals Alibaba's shift toward high-margin enterprise "AI Factories."