Hash Rate Ep. 191: TPN Subnet 65 on LLM Quantization-as-a-Service
markjeffrey · x · 2026-11-12
Hash Rate podcast Ep. 191 hosts the Mentor of TPN Subnet 65 (@tpnlabs/@GratisAI) covering LLM compression, 16-bit to 8-bit quantization, Fluid Bench and model-distribution scoring, why the subnet architecture, quantization-as-a-service, open vs. closed lab competition, and upcoming releases and benchmarks.
More from Infra
- Prosumer GPU thermal design mocked: venting 450W exhaust straight into the next card's intake — RexDouglass · 2026-11-12
- Bloomberg: Nvidia leadership crisis may push AI customers to Google TPUs, $50B sales by 2027 — dinabass · 2026-11-12
- PyTorch recaps Ascend Beijing meetup: Chinese open-weight models hit 45% of OpenRouter traffic — PyTorch · 2026-11-12
- Rumor: DeepSeek pretraining 2T-param V5 on Huawei Ascend 950 chips — Kyrannio · 2026-11-12
- What we learned building an AI token usage dashboard that actually helps — rseroter · 2026-11-12
- Cloudflare cuts Durable Objects cold starts and improves relocation routing, no code changes needed — irvinebroque · 2026-11-12