Browse the whole GLM model family on one page. The GLM line has widened quickly, and the releases now cover very different computational budgets: a 30B Flash tier built for cheap local inference, an Air tier tuned for performance per unit of compute, a speech-recognition model, and the 1M-context GLM-5.2 flagship built for long-horizon agentic engineering. Whichever GLM model you are evaluating, the goal of this hub is to get you to hard numbers quickly rather than to marketing adjectives.
Each model card links to a dedicated page with its spec strip, the benchmark data the model authors have published for it, a mini comparison chart against the closed-weight frontier and a short FAQ. Two things stay constant across the family: every core GLM model is released under an MIT license with weights published openly, and every benchmark figure quoted anywhere on this site is a result as published by the model authors, not something measured here. Where no figure has been published for a particular release, the page says so instead of estimating one — an honest gap is more useful in a comparison table than a plausible-looking invention.
The 1M-token context flagship. It is the open-source state of the art on coding and long-horizon evaluations, with explicit Non-Thinking, High and Max effort levels.
The previous flagship. It brought the first large long-horizon gains to the family and remains the baseline every GLM-5.2 improvement is measured against.
The release that opened the 5 series: stronger coding fundamentals, more reliable multi-step execution and noticeably better behaviour on complex agent tasks.
A 5-series base model tuned for dynamic, long-chain agent scenarios where the plan changes repeatedly and the trajectory has to stay executable.
A balanced reasoning model with enhanced programming capability, stable multi-step reasoning and markedly improved front-end generation.
The lightweight 30B tier. Efficient and fast enough for local inference, and reported to outrank open models of a similar scale.
The original efficiency milestone of the family: a small-footprint GLM model with strong performance for every unit of compute it consumes.
The dedicated speech model. Real-time speech-to-text recognition at a published character error rate of 0.0717.
Picking a GLM model comes down to two questions: how much context does the task genuinely need, and how much hardware do you have. Answer those two and the shortlist usually collapses to one or two candidates. For long-horizon software engineering, automated research loops and anything that has to keep an entire repository in view, GLM-5.2 is the only member of the family with a published 1M-token context window and explicit effort levels. If you are running on a single workstation GPU or need very high throughput on simpler reasoning, the lightweight GLM-4.7-Flash and GLM-4.5-Air tiers exist precisely for that constraint. If your workload is transcription rather than text generation, GLM-ASR is the dedicated speech model. Everything in between — GLM-5, GLM-5-Turbo and GLM-4.7 — represents earlier generations of the same reasoning line, useful when you want a smaller, more predictable model that still runs on open weights.
The table below summarises how the eight GLM model pages on this site line up. Where a figure has not been published by the model authors, the cell says so rather than guessing.
| Model | Category | Context | License |
|---|---|---|---|
| GLM-5.2 | Flagship reasoning | 1,000,000 tokens | MIT open source |
| GLM-5.1 | Reasoning | 200,000 tokens | MIT open source |
| GLM-5 | Reasoning | Not published | MIT open source |
| GLM-5-Turbo | Reasoning | Not published | MIT open source |
| GLM-4.7 | Reasoning | Not published | MIT open source |
| GLM-4.7-Flash | Lightweight (30B) | Not published | MIT open source |
| GLM-4.5-Air | Lightweight | Not published | MIT open source |
| GLM-ASR | Speech recognition | Audio streams | MIT open source |
All figures on this page are results as published by the model authors. glmmodel.com reports them; it does not run these evaluations.
Read the GLM model benchmarks, then put a real model to work. The playground above is free and needs nothing from you; if you want a fuller AI toolkit, our partner's free tier starts here.