Triangle Launches Cross-Model KV Cache Transfer, Up to 10.8x Faster TTFT at 1M Context
ycombinator · x · 2026-11-17
OneTriangle announced cross-model KV cache transfer: prefill on a small model, transfer the KV cache, decode on a large one — up to 10.8x faster time-to-first-token at 1M context. SDK is live for Qwen3.5-4B → Qwen3.8-27B, with beta open for more model pairs.
More from Infra
- Anthropic's annualized revenue hits ~$81B ahead of IPO, with $895B compute commitments — lennysan · 2026-11-17
- Sandbox startup speed is moot if compute sits 200ms from your infra — DanielLockyer · 2026-11-17
- PyTorch Foundation Honors 15 Winners in Fourth Annual Contributor Awards — PyTorch · 2026-11-17
- NVIDIA ops chief Deb Shoquist wins GSA's 2026 Woman of Innovation Award — nvidia · 2026-11-17
- Sam Altman walks back RAM comments: 'I want everyone to have a lot of RAM' — The Verge AI · 2026-11-17
- Chip Equipment Cycle Gets Longer: 2026 Wafer Fab Equipment Forecast Raised to $153B as Cleanrooms and ASML Output Cap Installations — BenBajarin · 2026-11-17