GLM 5.1

GLM 5.1 - Flagship Agentic Engineering Model

GLM-5.1 is Z.ai's next-generation flagship model for agentic engineering: an update to GLM-5 built specifically to stay effective on long-horizon coding and tool-use tasks where earlier models, including GLM-5 itself, tend to plateau.

Defining Characteristics

• Sustained Long-Horizon Optimization: revisits its own reasoning and revises strategy through repeated iteration, sustaining optimization over hundreds of rounds and thousands of tool calls — performance keeps improving the longer it runs.
• State-of-the-Art Agentic Coding: achieves SOTA performance on SWE-Bench Pro and leads GLM-5 "by a wide margin" on NL2Repo (repository generation) and Terminal-Bench 2.0 (real-world terminal tasks).
• DeepSeek Sparse Attention MoE: built on the GLM-5 architecture, which integrates DeepSeek Sparse Attention (DSA) to reduce deployment cost while preserving long-context capacity.

Why It Exists

Built to fix a specific failure mode in prior models, including GLM-5: exhausting familiar techniques early and plateauing on agentic tasks, regardless of how much time or compute they're given.