Qwen3.5-0.8B 4-bit spotted running on Rabbit R1 via wllama
ngxson · x · 2026-10-19
Developer mervenoyann reports she couldn't get llama.cpp running directly, so she used ngxson's wllama (a WASM binding of llama.cpp for the browser) to run a 4-bit quantized Qwen3.5-0.8B on the Rabbit R1 — an early public appearance of the unreleased "Qwen3.5" model name. ngxson quipped about the rabbit being "cooked," a jab at the R1's reputation.
Notable as a demo of running fresh small models locally on constrained devices via in-browser WASM inference.
More from Infra
- Nebius earns NVIDIA Exemplar Cloud validation for HGX B300 training workloads — DavidLinthicum · 2026-10-19
- RTX Spark on Surface Laptop runs Qwen 27B at 20.7 tok/s in Extended benchmark — draginol · 2026-10-19
- Getting compute in 2026: prepayments, SPVs and delayed power, per engineer — marksaroufim · 2026-10-19
- AI Toolkit adds LoRA training for Qwen 3.8 27B: 70-80 tok/s captioning under 20GB VRAM on RTX 5090 — ostrisai · 2026-10-19
- Local LLM on a gaming laptop: Qwen 27B hits 441 tok/s prefill but only 9.1 tok/s decode on RTX 5080 — draginol · 2026-10-19
- vLLM is now ~2x faster thanks to agentic-generated fused model definitions — mohbibi_ · 2026-10-19