Qwen 3.8 Max (Final Version Review & Free Ways): Okay, it's ACTUALLY a TOP MODEL!
In this video, I'll be telling you about the official launch of Qwen 3.8 Max, Alibaba's new 2.4 trillion parameter flagship model that beats Opus 4.8 on most benchmarks, costs a fraction of the price, and is going fully open weight next week. -- Key Takeaways: 🚀 Qwen 3.8 Max is now officially launched, a 2.4T parameter MoE model with a 1M token context window and native multimodality. 🔓 The full open weights drop next week, along with a smaller Qwen 3.8 27B model for local use. 🤖 Qwen showcased 10+ days of fully autonomous coding, from empty folder to production, with the complete trace published on GitHub. 📊 It tops PaperBench and Terminal Bench 2.1, beating both Opus 4.8 and Fable 5, and is the best computer use model on OSWorld-Verified. 🏆 On my KingBench, it holds second place overall at 81.25%, just behind Fable 5 and above Opus 4.8. 💸 API pricing is just $2/M input and $6/M output, with implicit caching at $0.25/M, a fraction of what closed labs charge. 🆓 You can use it for free on Qwen Studio, though the Qwen Code free tier remains discontinued. 👍 Overall, this is arguably the biggest open source AI moment yet, and the closed labs should be worried.
