Speculation Intensifies Over Alibaba’s Qwen 4 Release
Speculation regarding the release of Alibaba’s next-generation AI model, Qwen 4, has increased following unverified leaks and conflicting rumors circulating within developer communities. While official confirmation of a launch date remains absent, technical indicators and official statements from Alibaba representatives have provided a clearer picture of the project’s direction.
GitHub Leaks Suggest New Model Variants
Recent activity within the QwenLM GitHub repository has fueled technical speculation. According to unverified reports from Reddit users, pull requests in the qwen-code repository appear to reference several new model identifiers. These leaked strings include qwen4-max, qwen4-coder-plus, and qwen4-omni-120b.
Of particular interest to the developer community is the mention of a qwen4-omni-120b-think model, which is reportedly already being utilized as a session model in certain omni fixtures. If these identifiers are accurate, the “omni” designation suggests a multimodal capability, potentially allowing for verbal, back-and-forth interaction, a feature that has not been seen in the series since Qwen 3. However, these findings remain unconfirmed by Alibaba and may stem from testing or typographical errors.
Commitment to Open-Weight Distribution
Amidst rumors suggesting that Qwen 4 might mark a transition toward an API-only service model, Alibaba leadership has issued a clarification. Qwen Head Liu Dayi Heng stated that both Qwen 4.0 and Qwen 5.0 will continue to be released as open-source models.
Liu emphasized that the company intends to treat AI as a foundational capability accessible to a global user base rather than a proprietary tool restricted to a few entities. This stance follows the significant market footprint established by the Qwen series, which has surpassed 3 billion cumulative global downloads and includes over 460 open-source models and 300,000 derivative models.
Community Sentiment and Hardware Requirements
The anticipation surrounding the release has led to significant discussion on social media platforms. While some users have speculated on a release date as early as November 4, these claims lack official verification.
Technical discussions among users have also focused on the hardware implications of a potential 120B parameter model. Community members have noted that running high-parameter quantizations would require substantial system resources, with suggestions that 64GB of RAM may be necessary to run less “lobotomized” versions of large-scale models comfortably.