180B Qwen model runs on one DGX Spark: 2.39-bit quant keeps 95.5% of BF16 scores
TheZachMueller · x · 2026-10-02
Developer DJLougen released an extreme low-bit quantization of Qwen3.8-Flash-Next, now on Hugging Face.
- 2.125-bit experts, 2.39 bits/weight across the transformer
- 92 GB download, 39 GiB in memory
- Retains 95.5% of BF16's benchmark scores
The upshot: a 180B-parameter MoE model running on a single DGX Spark, with thanks to Lambda and HF's Zach Mueller for compute support.
Related event: 180B Model Squeezes onto a Single DGX Spark via 2.39-bit Quantization(2 posts)→
More from Infra
- tandem.cpp v0.3.0: 2-6x faster prefill for models too big for VRAM, Maya DeepSeek-V4.1 support — smflx · 2026-11-16
- Developers poll real costs of running coding agents on remote VMs for work — radim11 · 2026-11-16
- Datadog's new gateway returns 402 when a model's budget runs out—so agents just retry — YvesMulkers · 2026-11-16
- Max-quality local Minimax H3 video on a 4090: community-tested int8 setup — newaccount47 · 2026-11-16
- Local V2V outfit transfer on a 16GB M3 MacBook Air: 15.7 min cut to 2.4 min — might_thinker · 2026-11-16
- FlashStates hits 3.14x vLLM prefill throughput for hybrid-architecture agentic inference — 机器之心 · 2026-11-16