SenseNova Loop 9B on a single RTX 4080: 256k context for 16GB VRAM GPUs
roofkid · reddit · 2026-11-21
Reddit user roofkid added SenseNova Loop 9B support to his ninfer-4080 port, bringing a model claimed to match Qwen 3.8 27B with 256k context and vision to 16GB-VRAM RTX 4xxx cards.
- New convenience script pulls prebuilt Docker images and the runtime artifact — only prerequisite is Docker, no compiling or library setup
- The old script still runs Qwen 3.8 27B GSQ3 (100k context, text-only)
- The artifact is based on the AWQ quantization, which the author found scored higher on MBPP than quantizing from BF16 with ninfer's existing recipes
A practical way for low-VRAM users to get far more context while waiting for Qwen4 27B.
More from Infra
- SpaceX sells $32B of debt at 8%+ yields to secure Nvidia Rubin GPUs — rohanpaul_ai · 2026-11-21
- New US chip rules reportedly require licenses for non-US fabs and restrict overseas datacenters — JensHonack · 2026-11-21
- Portuguese tutorial walks through running MiniMax H3 locally on PC — Ashamed-Rub5601 · 2026-11-21
- Ex-NVIDIA staffer joins Gimlet Labs after NVIDIA's reported $20B bet on Groq's inference tech — yawnxyz · 2026-11-21
- Fully local pipeline: open-weight 27B models ship production code with Playwright + vision-loop tests — awitod · 2026-11-21
- Vespa co-founders launch Hornet: solving retrieval for the agentic era — jobergum · 2026-11-21