作者： Jakub Rusinowski · 最后更新： 2026年9月11日
The first natively multimodal model in the GLM-5 series, and the one Z.ai actually released under MIT — 320B total parameters with 18B active, text plus image and video in, a 1M-token context, and hybrid attention (linear attention for local dependencies, sparse attention for distant retrieval). Z.ai claims 63.4 on DeepSWE against GLM-5.2's 46.2. Note this is NOT the GLM-5.3 flagship: that is a separate 753B model under a bespoke licence with a hyperscaler security-review clause.
| Licence | What it permits | Applies to |
|---|---|---|
MIT | Commercial use permitted Commercial use permitted. No usage restrictions beyond attribution. | GLM-5.3-Flash 320B-A18B |
| GLM-5.3-Flash 320B-A18B | Min 194 GB VRAM · Q4_K_M · 1,048,576 ctx · ollama run glm-5.3-flash:cloud |
The cheapest GPU that runs GLM-5.3-Flash locally (min 194 GB VRAM) is the Apple M3 Ultra (512 GB).
Install Ollama then run: ollama run glm-5.3-flash:cloud
Minimum VRAM: 194 GB. For best results use Q4_K_M quantization.
GLM-5.3-Flash needs about 194 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: GLM-5.3-Flash 320B-A18B (194 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
GLM-5.3-Flash's smallest variant needs about 194 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
Q4_K_M is the best balance of quality and VRAM for GLM-5.3-Flash in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
Install Ollama, then run: ollama run glm-5.3-flash:cloud. This downloads GLM-5.3-Flash and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.