14 Days to Break 1 Million Downloads! Zhipu GLM-4.7-Flash Leads the Open Source Large Model SOTA

Zhipu AI officially announced that its latest open-source lightweight model GLM-4.7-Flash has already exceeded 1 million downloads on the international mainstream open-source community Hugging Face just two weeks after its release.

As a hybrid thinking model with 30B-A3B, GLM-4.7-Flash demonstrates strong performance. According to the latest benchmark test data:

In mainstream tests such as SWE-bench Verified and τ²-Bench, its overall performance has surpassed gpt-oss-20b and Qwen3-30B-A3B-Thinking-2507.
In models of the same size and similar sizes, this model has successfully achieved the open-source SOTA (State-of-the-Art) score.

Zhipu AI stated that GLM-4.7-Flash is designed to provide global developers with a lightweight deployment option that balances reasoning efficiency, practical application performance, and high cost-effectiveness. The achievement of a million downloads marks a high level of recognition for this model in the open-source ecosystem, further promoting the popularization and application of large model technology.

Jan Team Releases Jan-v2-VL-Max! A 30B Multimodal Model Specializing in Long-Term Agent Tasks, Stable Execution of Long Sequences Outperforms Gemini 2.5 Pro

The Jan Team releases the 30B-parameter multimodal large model Jan-v2-VL-Max, designed for long-term and highly stable automation tasks. Its performance surpasses Google's Gemini 2.5 Pro and DeepSeek R1. The model focuses on solving the 'error accumulation' and 'loss of focus' issues in multi-step tasks, providing strong support for the open-source agent ecosystem.

Fedora Community Releases Draft of AI-Assisted Contribution Policy for Feedback

Fedora community released the draft of "AI-Assisted Contribution Policy" and initiated a two-week feedback period. The policy aims to regulate the use of AI technology and ensure that open source values are not compromised. The draft covers key areas such as overall principles, emphasizing the maintenance of community core values. The council will review and vote on whether to officially adopt it.

14 Days to Break 1 Million Downloads! Zhipu GLM-4.7-Flash Leads the Open Source Large Model SOTA

Related Recommendations

Only 20% of the quota is released daily! Zhipu AI Announcement: GLM Coding Plan will start limited sales from January 23rd

New Benchmark for 30B Specifications! Zhipu AI Open Sources GLM-4.7-Flash, Outperforming Alibaba and OpenAI in Multiple Tests

Jan Team Releases Jan-v2-VL-Max! A 30B Multimodal Model Specializing in Long-Term Agent Tasks, Stable Execution of Long Sequences Outperforms Gemini 2.5 Pro

Generate a Professional PPT in One Sentence! Open Source Tool Presentation-AI Now Available for Free, Runs Locally, Outperforms Gamma.app

Fedora Community Releases Draft of AI-Assisted Contribution Policy for Feedback