GLM 5.3 Flash runs locally with Blender/Unity CLI access
antirez · x · 2026-09-01
This post showcases GLM 5.3 Flash Q2 running locally with access to Blender and Unity CLIs. It uses ds4 with vision support on an M5 Max (128GB). The test achieved an average speed of 17.32 t/s with a context length of 120128 on Auto power mode.
More from Infra
- Energy, Grid, Cooling First: Why AI's Real Competitive Stack Starts Below Compute — ingliguori · 2026-09-01
- Comet releases Opik, an open-source LLM observability tool for debugging and monitoring — dl_weekly · 2026-09-01
- VMware Expert: Private AI Offers Economic and Compliance Edge — DavidLinthicum · 2026-09-01
- Obscura: A Lightweight Rust Browser Built Exclusively for AI Agents — Shruti_0810 · 2026-09-01
- RTX 3090 runs Qwen2.5-72B at 2,000 tok/s prefill — iamMess · 2026-09-01
- Space Data Centers Cost 20x More to Launch Than to Build on Earth — aronchick · 2026-09-01