Zhipu's GLM 3.5 Flash Served 42T Tokens in 6 Days Free on Chinese Chips
bindureddy · x · 2026-08-27
A breakdown of how Ox-Alpha (GLM 3.5 Flash) processed 42 trillion tokens for free over 6 days, entirely on Chinese silicon.
Key points:
- Custom SGLang inference engine with disaggregated Encode-Prefill-Decode architecture
- A GLM-5.3 infrastructure agent wrote its own GPU kernels and debugged bottlenecks — the model optimized its own serving stack
- 3× end-to-end performance gains, Nvidia-level cost parity on domestic silicon
- Backed by Zai's 1GW data center: 10,000+ chip clusters, zero Nvidia silicon
The author argues export controls backfired, forcing a fully sovereign Chinese AI stack that works.
More from Infra
- SGLang ships Day-0 support for FLUX 3 Image Klein: 10 reference images, BF16/FP8/NVFP4 variants — ying11231 · 2026-10-16
- NVIDIA's Nemotron 3.5 Super VL hits CoreWeave: 12B-active multimodal MoE at $0.25/M input tokens — wandb · 2026-10-16
- Chip equipment beats bull case: 2026 WFE seen at $153B, $278B by 2030 — BenBajarin · 2026-10-16
- Steam now launches in 5.23s, beating Epic's 5.47s — and vLLM gets roasted for slow startup — AlpinDale · 2026-10-16
- Shadeform to Announce Six New Features on Oct 19: Lower GPU Costs, More Suppliers — _jaydeepkarale · 2026-10-16
- Fastino launches Training API for SFT/RL, claiming 25-85% cost cut vs Tinker — kalyan_kpl · 2026-10-16