Run GLM 4.6 locally
GLM 4.6 is 357B parameters (32B active per token — a mixture-of-experts, so it still needs room for all 357B in memory). Here's the VRAM it needs at each quantization, the smallest GPU that fits, and how a self-hosted box compares with the cheapest verified API.
VRAM by quantization
| Quantization | VRAM needed | Fits on | Notes |
|---|---|---|---|
| int4 (Q4) | 214.2 GB | 4× H100 (320 GB) ($6.92/hr) | smallest footprint, minor quality loss |
| int8 (Q8) | 428.4 GB | 8× H100 (640 GB) ($13.84/hr) | near-lossless |
| fp16 (full) | 856.8 GB | multi-node (>640 GB) | reference quality |
VRAM ≈ params × bytes/param × 1.2 (overhead). See the fullmethod and break-even calculator.
Self-host vs API
Cheapest API for GLM 4.6 is $2.18 / 1M tokens(blended) via Venice. Running it yourself in int4 fits a4× H100 (320 GB) at $6.92/hr — about $5,051.6/month at 24/7. Those two lines cross at roughly 2.3B tokens/month: below that the API wins on cost, above it the dedicated GPU does (assuming you keep it busy). Tune your own volume in the break-even calculator.
Run it at home
No cloud account needed: GLM 4.6 in int4 (214.2 GB) fits a Mac Studio M5 Ultra (256 GB unified), a $9,499 one-off buy. Amortized over 3 years that's about $0.3615/hr — hardware only, electricity aside. Against the Venice API at $2.18/1M, buying it pays for itself after roughly 4.4B tokens total. Below that the API is cheaper; a machine that mostly sits idle rarely earns back its price.
Get the weights
Quantized builds on the model hubs — links search live, so they track new builds as they appear:
- GGUF builds llama.cpp · Ollama · LM Studio
- AWQ builds GPU serving (vLLM)
- GPTQ builds GPU serving
- MLX builds Apple silicon
- Ollama library one-command local run
License: open weights under MIT · terms. Compare all providers on the glm-4.6 API page.