Leaked GLM-5.5 Rumored for August Release with 1T+ Parameters and Open Weights

Recent online leaks indicate that the GLM model series may skip version 5.3 and directly launch GLM-5.5 in August (within 4 weeks). This development is drawing significant attention, as the rumored new model is highly competitive in parameter scale, context capabilities, and openness, directly targeting the most cutting-edge AI labs.

Key Specs and Features

Based on the leaked information, GLM-5.5 could have over 1 trillion total parameters. The model will continue to feature open weights, allowing users to acquire and deploy it locally, while retaining a 1 million token context window. In terms of capabilities, GLM-5.5 will primarily focus on long-running autonomous coding and agentic tasks.

Goals and Teaser Dynamics

Performance-wise, leaks suggest that GLM-5.5 aims to directly compete with top-tier companies like OpenAI, Anthropic, Google, and xAI, with expectations to surpass Fable 5. Meanwhile, there have been hints at the official level: a tweet stating "now might be the best time to subscribe to the GLM Coding Plan" is viewed by the public as a restrained teaser for the upcoming model.

2026-07-20 ~ 2026-07-21 · 5 related posts

Primary sources