GLM-4.6

by Z.ai (GLM project)

Open-source flagship coding and reasoning model with an expanded 200k context window

See vendor docs and community benchmarks

Summary

GLM-4.6 is a 2025 open-source model release focused on coding, agentic workflows, and expanded-context reasoning. It improves on GLM-4.5 with a larger context window (~200k tokens), better coding performance, and improved token efficiency.

Features

  • 200k token context window (large-context workloads)
  • Strong coding performance and multi-file refactor capabilities
  • Improved reasoning and agentic tool use
  • Integrations with coding platforms and CLIs (GLM Coding Plan, OpenRouter)
  • Competitive performance vs Deepseek V3.2 and Claude Sonnet models in benchmark suites

Superpowers

GLM-4.6 is for developers and research teams wanting an open-source model that handles large codebases and long-running agent tasks with better token efficiency and strong coding outputs.

Pricing & access

  • Available via GLM Coding Plan subscriptions and self-hosting through community distributions. Check Z.ai and OpenRouter for current access paths and pricing.

Known limitations & notes

  • May trail some closed-source models (e.g., Sonnet 4.5) on certain coding benchmarks
  • Self-hosting large-context models requires substantial infra (GPU RAM, NVMe)

Sources / notes:

  • Community benchmarks, release notes, and comparative analyses.