unsloth/Qwen3.8-27B-NVFP4
published by Unslothhuggingface.co/unsloth/Qwen3.8-27B-NVFP4
This is a quantization of Qwen3.8 27B, not an independent model. Parameters, benchmarks, prices and lineage are recorded on the canonical model. Open Qwen3.8 27B →
Updated 3 h ago · first seen 11 Sept 2026
model_01M2950GTKRA4TT6325FC8AKVA
- File size
- Format
- Downloads
- Published
Artifact facts
- Weights dtype
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- File size
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- Downloads
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 6 h agomedium
- Likes
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 6 h agomedium
- Hugging Face repo
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium
- Base model
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- Gated
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- License
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- Release date
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- Last modified
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium
- Model card
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium
- Tags
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
Hardware fit (this packaging)
Assumptions (7)
- Estimated, not measured: weights = parameters × bytes/param × 1.15 runtime overhead (or the observed artifact file size when one is recorded).
- bytes/param: 4bit = 0.5, 8bit = 1.0, fp16 = 2.0 (uniform quantization, no per-layer exceptions).
- KV cache: 2 × layers × kv_heads × head_dim × 2 bytes × context × batch when the architecture is known; otherwise 0.5 GB per 8 192 tokens (× batch), independent of architecture (GQA/MLA models need less).
- A model 'fits' when the estimate is at most the device memory minus 2 GB reserved for the OS and framework.
- Mixture-of-experts models are estimated on total parameters (all experts must be resident); active parameters are ignored.
- Device memory uses the largest configuration when several are listed (e.g. Apple silicon tiers).
- Multi-GPU: device memories are summed; interconnect bandwidth, tensor-parallel replication and pipeline bubbles are not modelled.
Provenance
Attributed facts
22
Source tiers
T222
Freshest observation
3 h ago
Conflicts
None
Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.