ByteDance has pushed back Doubao 2.2, which was supposed to launch in August, and plans to spend extra time on training, with coding, tool calling, and agent capabilities as priorities. According to BlockBeats, one of Seed's targets this year is to build a first-tier coding model. Internally, ByteDance hopes to reach a level similar to GLM-5.2 and Kimi-K3 by year-end, so developers begin to trust its coding model. The delay follows an August reorganization at Seed. Teams previously divided by text, speech, code, and vision have been recast as Pretrain Data, Horizon RL, Product Posttrain-Work, and Product Posttrain-Chat. Horizon RL will handle coding post-training; Work will concentrate on tool calling, GUI operations, and long task execution. ByteDance is also recruiting aggressively. Open positions include Code Agent, general Agent, and reinforcement learning algorithm engineer. A Multi-Agent Harness posting explicitly mentions studying Claude Code, Codex, and other coding agents, and building multi-agent plus reinforcement learning environments designed to run for long stretches. Seed 2.1 already highlighted coding when it came out in June, but ByteDance appears to think that was insufficient. Doubao 2.2 is being delayed intentionally, so coding can be upgraded to a more credible level before release.
ByteDance Pushes Back Doubao 2.2
ByteDance has delayed Doubao 2.2, a model originally scheduled for an August launch, according to BlockBeats. The company plans to spend more time on training and wants to fix three areas first: coding, tool calling, and agent capabilities.
Making its coding model genuinely strong is one of Seed's goals this year. Internally, ByteDance wants a model close to GLM-5.2 and Kimi-K3 by the end of the year, with enough quality that developers start to take its coding model seriously.
Seed Reorganizes Into Four Teams
The delay follows a big internal reshuffle at Seed in August. Teams that used to be split by text, speech, code, and vision have been broken apart and regrouped into four units: Pretrain Data, Horizon RL, Product Posttrain-Work, and Product Posttrain-Chat.
Under the new structure, Horizon RL owns coding post-training. The Work team is responsible for tool calling, GUI operation, and long-task execution.
Recruiting Points to Claude Code and Codex
ByteDance has been hiring heavily in related roles. Openings on its site include Code Agent, general Agent, and reinforcement learning algorithm engineer. One Multi-Agent Harness posting goes further, saying the role will study coding agents such as Claude Code and Codex and build multi-agent and reinforcement learning environments that can run continuously for a long time.
Seed 2.1, released in June, already put coding at the center. ByteDance clearly does not think that was enough. Doubao 2.2 can ship late, but the coding capability needs to look credible first.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
