Zhipu's GLM-5.3-Flash Served Record Traffic Entirely on Domestic Chinese Chips
teortaxesTex · x · 2026-08-27
Zhipu's GLM-5.3-Flash — the 320B-A18B model revealed this week as the anonymous "Ox Alpha" — set usage records on OpenRouter and OpenCode, and the company says all of that traffic was served by domestic Chinese chip clusters.
teortaxesTex notes the striking part: these aren't even China's best chips or the current mass-production generation, and the situation should improve by end of year. Huawei 950s were reportedly slated for mass sales in August. A Zhihu contributor reconstructed how Zhipu lined up the infrastructure and read official engineering details for hints on which chips are doing the work. The core takeaway: the bigger story is a frontier-level model handling global-scale real-world inference on domestic silicon, not the model itself.
More from Infra
- Test: OX Alpha runs on WebGPU for just $0.016 — yuwen_lu_ · 2026-08-27
- Local AI trade-off: 96GB mixed RAM vs. speed — QuirksNFeatures · 2026-08-27
- PyTorch Ecosystem Adds Perforated, TokenSpeed, and 8 Others — zhyncs42 · 2026-08-27
- Jalapeño chip delivers up to 1.9x efficiency, docs criticized — Artistic_Phone9367 · 2026-08-27
- 8x RTX 3090 Setup Serves Qwen Flash Next at 661 tok/s with 262k Context — QuixiAI · 2026-08-27
- OpenAI Internal Compromise Deemed More Critical than Hugging Face Incident — sjgadler · 2026-08-27