Alibaba's Qwen AI model achieved 33 rounds of self-improvement without human intervention, boosting its intelligence score. This marks a significant step towards more autonomous AI development and massive future models.
Big news from the AI world! Alibaba recently announced a truly impressive achievement for their Qwen AI model at the Apsara Conference. Imagine this: their Qwen3.8-Max model managed to train and refine itself through 33 continuous rounds over more than a month, all without any human help! This means AI is now learning and improving its own capabilities, opening the door to a very different future.
The result of this self-evolution was quite remarkable: its Artificial Analysis Intelligence Index score jumped from 40 to 45. What's even more fascinating is that the core model code didn't change; instead, its underlying capabilities improved significantly. This means users calling the old model by the same name are now getting a smarter, improved version, thanks to a completely autonomous learning process.
So, what does this mean for you? Firstly, it's a glimpse into AI's growing ability to become more independent in learning and self-improvement. Secondly, Alibaba isn't stopping there; they've announced ambitious plans to develop Qwen 4.5 and Qwen 5 models, aiming for a massive scale of 5 to 10 trillion parameters. This huge size implies vastly greater comprehension and processing power.
While these achievements are Alibaba's statements and haven't been independently verified yet, the practical example they shared was quite compelling. They described how the model created a command-line tool called «oh-my-cli» from scratch. The model initiated a self-development process, building the tool, testing it, and whenever it encountered an error, it would go back and fix it on its own, in a continuous loop of refinement. In just 16 days, the model added 265 commits to the code. This shows us how AI could become an independent software engineer, capable of creating and debugging efficiently. We're truly seeing the beginning of a new era where human intervention might not be needed at every step of AI development.
The result of this self-evolution was quite remarkable: its Artificial Analysis Intelligence Index score jumped from 40 to 45. What's even more fascinating is that the core model code didn't change; instead, its underlying capabilities improved significantly. This means users calling the old model by the same name are now getting a smarter, improved version, thanks to a completely autonomous learning process.
So, what does this mean for you? Firstly, it's a glimpse into AI's growing ability to become more independent in learning and self-improvement. Secondly, Alibaba isn't stopping there; they've announced ambitious plans to develop Qwen 4.5 and Qwen 5 models, aiming for a massive scale of 5 to 10 trillion parameters. This huge size implies vastly greater comprehension and processing power.
While these achievements are Alibaba's statements and haven't been independently verified yet, the practical example they shared was quite compelling. They described how the model created a command-line tool called «oh-my-cli» from scratch. The model initiated a self-development process, building the tool, testing it, and whenever it encountered an error, it would go back and fix it on its own, in a continuous loop of refinement. In just 16 days, the model added 265 commits to the code. This shows us how AI could become an independent software engineer, capable of creating and debugging efficiently. We're truly seeing the beginning of a new era where human intervention might not be needed at every step of AI development.
