GLM-5.1

GLM-5.1 is positioned by Z.ai as a next-generation flagship model focused on agentic engineering, with coding capabilities that improve meaningfully on its predecessor GLM-5. The release targets complex software engineering workflows where the model needs to decompose problems, run experiments, read results, and identify blockers across extended sessions rather than producing a single quick answer. The model is shared as open weights, making the architecture and weights openly accessible for self-hosting and downstream research.

On complex software engineering benchmarks, GLM-5.1 is reported to reach state-of-the-art performance on SWE-Bench Pro with a score of 58.4, while also leading GLM-5 by a wide margin on NL2Repo repository generation and Terminal-Bench 2.0 real-world terminal tasks. Qualitative evaluations described by Z.ai highlight better judgment on ambiguous problems and sustained productivity over longer agentic sessions, with the model revisiting its reasoning and revising strategy through repeated iteration rather than plateauing after early gains. Practical fit centers on long-horizon coding agents, repository-scale generation, and terminal-driven development where the model must stay effective across many steps.

SCNet Token PlanGLM-5.1glm

Open provider page Provider docs

Quick Info

Provider: SCNet Token Plan
Model key: GLM-5.1
Release date: Apr 7, 2026
Last updated: Apr 7, 2026
Input modalities
Output modalities
Capabilities

Cost

A provider subscription or plan supersedes token-based pricing for this model.

Limits

Output tokens: 131,072 tokens
Context window: 200,000 tokens

Latest news about GLM-5.1

SCNet Token Plan

CoverageBenchmark

MorphLLM's technical breakdown covers GLM-5.1 as Z.ai's 744B-total / 40B-active mixture-of-experts coding model released April 7, 2026 under the MIT license, with a 200K-token context window and up to 128K-131K max output. It documents that GLM-5.1 was a post-training upgrade to the GLM-5 base model, delivering roughly The article records GLM-5.1's launch benchmarks — 58.4 on SWE-bench Pro (topping GPT-5.4 at 57.7 and Claude Opus 4.6 at 57.3) and 63.5 on Terminal-Bench 2.0 — and notes the model is engineered for up to 8-hour autonomous agentic runs. It also details runtime compatibility with transformers, vLLM, SGLang, KTransformers,

SCNet Token Plan

CoverageRelease Notes

Z.AI's official release notes document the GLM-5.1 model with an entry dated 2026-04-07, positioning it as a long-horizon, agentic upgrade in the GLM family. The page states GLM-5.1 is designed for sustained autonomous work — capable of running independently for up to 8 hours in a single session — covering planning, ex Because this is the creator's own release-notes page, it also serves as the attribution anchor: GLM-5.1 is a Z.AI model, and the surrounding entries confirm a rapid family cadence with GLM-5.2 (2026-06-16) and GLM-5.3 (2026-08-18) following it. That sequencing matters for any reader using the notes to understand where