Alibaba Unveils 2.4 Trillion-Parameter Open-Source Large Model, Challenging Fable 5
The open-source community has recently achieved a major milestone with the official release of Qwen3.8-2.4T-A95B model weights by Alibaba’s Qwen team. This marks the first time the team has fully opened its flagship Max-level model to global developers and researchers.
At the core of this release is a model with 2.4 trillion total parameters, activating 95 billion per token. It natively supports an ultra-long context window of 260,000 tokens, expandable to 1,010,000 tokens. Building on the Qwen3.5 architecture, it delivers comprehensive performance gains in coding, office productivity, research assistance, and long-term autonomous agent tasks.

In authoritative benchmarks, the model demonstrates top-tier performance, ranking highly on PaperBench for paper reproduction and OSWorld-Verified for computer operation capabilities. It even outperforms several international mainstream models in key metrics. Notably, its robust long-term autonomy allows it to independently handle complex programming tasks spanning days and autonomously reproduce or optimize research papers.
Related article
Anthropic's Claude Used to Breach OpenAI
In a development that underscores the evolving landscape of AI security, independent researchers leveraged Anthropic’s Claude to breach OpenAI, revealing vulnerabilities in the ChatGPT developer’s defenses, as reported by The Wall Street Journal on T
ByteDance Delays Launch of Douyin Large Model 2.2
ByteDance has postponed the August release of the DouBao large model 2.2 to allow for a more comprehensive enhancement of its capabilities in coding, tool invocation, and Agent intelligence. This strategic delay aims to deliver significant performanc
Google Invests $10M in Bankrupt Airline, Prioritizing AI and Product Development
Following a recent ruling by the U.S. Southern District Bankruptcy Court, Google has successfully acquired key corporate assets from the defunct budget carrier Spirit Airlines for $10 million. In a competitive bidding war, Google outmaneuvered Mercor
Related Special Topic Recommendations
Comments (0)
0/500
The open-source community has recently achieved a major milestone with the official release of Qwen3.8-2.4T-A95B model weights by Alibaba’s Qwen team. This marks the first time the team has fully opened its flagship Max-level model to global developers and researchers.
At the core of this release is a model with 2.4 trillion total parameters, activating 95 billion per token. It natively supports an ultra-long context window of 260,000 tokens, expandable to 1,010,000 tokens. Building on the Qwen3.5 architecture, it delivers comprehensive performance gains in coding, office productivity, research assistance, and long-term autonomous agent tasks.

In authoritative benchmarks, the model demonstrates top-tier performance, ranking highly on PaperBench for paper reproduction and OSWorld-Verified for computer operation capabilities. It even outperforms several international mainstream models in key metrics. Notably, its robust long-term autonomy allows it to independently handle complex programming tasks spanning days and autonomously reproduce or optimize research papers.
Anthropic's Claude Used to Breach OpenAI
In a development that underscores the evolving landscape of AI security, independent researchers leveraged Anthropic’s Claude to breach OpenAI, revealing vulnerabilities in the ChatGPT developer’s defenses, as reported by The Wall Street Journal on T
ByteDance Delays Launch of Douyin Large Model 2.2
ByteDance has postponed the August release of the DouBao large model 2.2 to allow for a more comprehensive enhancement of its capabilities in coding, tool invocation, and Agent intelligence. This strategic delay aims to deliver significant performanc
Google Invests $10M in Bankrupt Airline, Prioritizing AI and Product Development
Following a recent ruling by the U.S. Southern District Bankruptcy Court, Google has successfully acquired key corporate assets from the defunct budget carrier Spirit Airlines for $10 million. In a competitive bidding war, Google outmaneuvered Mercor





Home






