AI Models and Office Agents Roll Out in Cluster: Qwen Office 2.0 Shifts to Team Agent

On Nov 10, QbitAI reported a cluster of updates on AI models and agent products, spanning office collaboration, image editing, multimodal creation and agent self-evolution. The items include: Qwen Office 2.0's shift to Team Agent, Seedream 5.5's faster and cheaper image editing, Fable 5.2's first hands-on tests covering composing, filmmaking and 3D modeling, and openJiuwen's release of the TTSE experience-learning framework.

Qwen Office 2.0: From Standalone Mode to Team Agent

According to QbitAI, Qwen Office 2.0 targets Team Agent, ending the standalone mode of office AI. The QbitAI team ran a full day of editorial work through the product to test its collaboration capabilities in real scenarios.

The report summary says "office AI ends the standalone mode," pointing to a new multi-agent collaborative office format.

Seedream 5.5: Faster Image Editing at a Lower Price

Seedream 5.5 brings updates to image-editing tasks. QbitAI reported that the version performed steadily in a "100-round image editing" test, with improvements to both speed and API pricing — faster speed and a lower API price.

The summary puts it plainly: "speed is up, API price is down."

Fable 5.2: Hands-On Tests Cover Composing, Filmmaking and 3D Modeling

The first hands-on tests of Fable 5.2 show the model can handle composing, filmmaking and 3D modeling, which QbitAI described as "a full one-stop package." Some users remarked that "Opus suddenly lost its appeal," reflecting the attention drawn by the model's competitiveness in multimodal creation.

openJiuwen: First to Launch the TTSE Experience-Learning Framework

openJiuwen released the TTSE experience-learning framework, positioned to enable agent self-evolution. QbitAI reported that the framework aims to "make every task the starting point of the next improvement," marking agent self-evolution's entry into a "dual-track era."

What to Watch

The products above come from different teams and cover subfields including office collaboration, image editing, multimodal generation and agent learning frameworks. Their actual performance, usability and market feedback remain to be seen.