{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"AI News Today | Julian Goldie Podcast","title":"China’s Qwen 3.7 Max DESTROYS Claude?","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/3b39a837\"></iframe>","width":"100%","height":180,"duration":912,"description":"Qwen 3.7 Max: Alibaba’s 35-Hour Autonomous Agent Demo + Claude-Beating Benchmarks (With Caveats)\nAlibaba’s new flagship model, Qwen 3.7 Max, was unveiled around the Alibaba Cloud Summit in Hangzhou (May 20, 2026) and is positioned as a closed, proprietary frontier model aimed at enterprise, narrowing the gap with Claude Opus 4.7 while costing less per token. The script highlights strong agentic benchmarks (e.g., Terminal Bench 2.0, SWE-Bench Pro, MC Atlas, GPQA Diamond) and broad compatibility with agent frameworks and APIs (OpenAI and Anthropic specs), plus availability across multiple platforms. It also stresses caveats: the model is unusually verbose, which can raise real costs, and it has a low hallucination rate partly due to a much lower attempt rate. A headline 35-hour autonomous optimization demo (vendor-stated, not independently verified) reportedly achieved a 10× speedup on Alibaba’s Shenwu M890 chip kernel.\n00:00 Qwen Shocks The Frontier\n01:34 What Qwen 3.7 Max Is\n02:17 Agent Framework Compatibility\n02:55 Benchmark Wins Explained\n04:12 Pricing And Token Trap\n05:29 Hallucinations Versus Refusals\n06:27 Inside The 35 Hour Demo\n08:19 Hermes Agent Integration\n09:50 Should You Switch Now\n11:39 How To Test And Deploy\n12:36 Stop Waiting Start Building\n14:31 Final Takeaways And Caveats","thumbnail_url":"https://img.transistorcdn.com/Tp0JSUlyiAe2ran44b13cjk4ImQYS1QEMKCAa9Q7go0/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9lNTYz/NGQxMTU2ZTM1MzU5/MDNiYjcwZjJmYTY5/ODJmOS5qcGc.webp","thumbnail_width":300,"thumbnail_height":300}