ModelsGLM-4.5V Guide: Grounding, GUI Agents and a 16K Output Cap
GLM-4.5V is Z.ai's vision-language model built on GLM-4.5 Air. How its grounding output works, where the 16K output cap bites, and when GLM-4.6V replaces it.
ReadNotes on models, agents, and what inference costs
Practical writing for developers building with large language models — how they work, how to pick one, and how to keep the bill predictable.
Picking between models — and testing them on your own work.
Models
ModelsGLM-4.6 Guide: 200K Context and Fewer Tokens Per Task
GLM-4.6 kept GLM-4.5's price and raised the context to 200K with a claimed 30% token saving. What changed, how to verify it, and where GLM-4.7 now fits.
Read
ModelsGLM-4.6V Guide: Function Calling With Images and 128K Documents
GLM-4.6V adds native function calling with images as tool inputs and reads long documents as pages. What it does, its known weak spots, and what it costs.
ReadModelsGLM-4.7 Flash Guide: A 30B-A3B Agent Model You Can Run Locally
GLM-4.7 Flash is a 30B-total, 3B-active MoE with MIT weights, a 200K context and a free tier on Z.ai. Its benchmarks, local setup, and when to step up.
ReadModelsGLM-4.7 Guide: The Last GLM-4 Flagship, and Where It Still Fits
Z.ai's December 2025 flagship is a 358B MIT-licensed model with a 200K window and three thinking controls. What it does, what it costs, and when to move up.
ReadModelsGLM-5 Guide: The MIT-Licensed Base of Z.ai's 5.x Family
GLM-5 grew to 744B total and 40B active parameters, added sparse attention and shipped MIT weights. What it is, what it costs, and when it still fits.
Read
ModelsGLM-5 Turbo: Z.ai's Agent-Tuned Model for OpenClaw Workflows
GLM-5 Turbo is an OpenClaw-tuned variant, not a cheaper GLM-5. What Z.ai says it changed, what it costs, why there are no weights, and how to test the claims.
Read
ModelsGLM-5.1 Guide: The Eight-Hour Agent Claim, Six Months On
GLM-5.1 claimed the top SWE-Bench Pro score at launch and eight-hour autonomous runs. What it is, what the 200K window costs you now, and when it is still worth running.
ReadModelsGLM-5.3 Flash Guide: 18B Active, Multimodal and MIT-Licensed
GLM-5.3 Flash is a new 320B hybrid-attention model with native image and video input, close to full GLM-5.3 on agent benchmarks at a fraction of the per-token price.
ReadModelsGLM-5.3 FlashX Guide: Paying 2.5 Times More for Speed
FlashX is the high-speed tier of GLM-5.3 Flash, quoted at up to 200 tokens per second for about 2.5 times the price. When that speed earns its cost and when it does not.
ReadModelsGLM-5.3 Guide: A Post-Training Release With Breaking Changes
GLM-5.3 keeps the GLM-5.2 base and gains everything from post-training. The benchmark jump, the always-on reasoning change, and the licence to read before shipping.
ReadModelsGLM-5.3 Prime Guide: Double the Price for Faster Full GLM-5.3
GLM-5.3 Prime is listed as full GLM-5.3 with 1.5 to 2 times the output throughput at twice the per-token rate. How to work out whether that trade pays for your workload.
ReadShowing 73–84 of 446 articles