qwegnai/SenseNova-Loop-9B-hq4g32
image-text-to-text model by qwegnai
From the publisher
Model card & documentation
Source previewRead the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.
A native weight pack for the Qwegn inference engine (v0.27.0+), a Swift 6.2 + Metal 4 runtime for Apple Silicon. HQ4G32 Hadamard block quantization paired with preserved F32 two-stream Sinkhorn-20 mHC / Gated-DeltaNet state planes and a Spherical Hadamard SHVQ4 Latent Sparse Attention sidecar (model.qlsa) packs sensenova/SenseNova-Loop-9B into 7.38 GiB (7,922,099,956 bytes, 4.66 GiB GPU trunk weights, 4.87 GiB server text RSS), delivering 2,044.9 tok/s prefill @4K (1,943.5 @16K, 1,763.2 @64K), 165.7 tok/s weight-free .ngram speculative decode (5.49 tok/step), 152.96 tok/s 4-stream HTTP decode (3.02×), 18 ms warm prefix cache (46.1×), 211/211 (100%) API conformance, 92% HumanEval (36/50 Code-50), and 3/3 needle recall at 16K and 64K with 0 steady-state allocations. Measured on an idle Apple M5 Max (Mac17,6, 18-core CPU, 40-core Apple10 NAX GPU,…
Read the full model card ↗ · Preview checked 2026-11-18T08:49:32.028Z
Inside the original model card — Document outline
- SenseNova-Loop-9B · HQ4G32 · Qwegn native pack
- At a glance
- Performance on Qwegn (v0.27.0)
- Quantization Layout
- Memory Footprint (qwg plan, 0 steady-state allocations)
- Quick start
- Requirements
- Files
- License and attribution
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Links from the model card
References supplied by the publisher, not independently verified endorsements. Check the destination before downloading files or entering credentials.
Documentation belongs to its respective authors. Reported project/model license: apache-2.0. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
qwegnai/SenseNova-Loop-9B-hq4g32 is a image-text-to-text model repository published by qwegnai on Hugging Face. The source reports the qwegn library.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- image-text-to-text
- Library
- qwegn
- Recent downloads (30 days)
- 0
- Cumulative likes
- 0
- Hugging Face trending score
- 0
- Reported safetensors parameters
- Not reported
- Architecture
- Not reported
- License
- apache-2.0
- Access
- Not gated by Hugging Face
- Created
- 2026-11-18T06:44:39.000Z
- Last modified
- 2026-11-18T06:47:24.000Z
Compatibility and lineage
Reported languages: en, zh
Reported base models: sensenova/SenseNova-Loop-9B
Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.
Use and evaluate
Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.
No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.
Source and freshness
Source: Hugging Face Hub. Metadata observed 2026-11-18T07:01:26.893Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.