← back to catalog · registered 2026-11-18 12:04

balllvin/SenseNova-Loop-9B-abliterated-arete-NVFP4-MLX-4bit

Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/balllvin%2FSenseNova-Loop-9B-abliterated-arete-NVFP4-MLX-4bit"
Response includes
  • classification m1
  • files 14
  • author_summary 28 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · 30-day
0
Likes
0
Model age
4d ago
created 2026-11-18

Training datasets

Corpora the author lists in the model card. Datasets tracked in our /datasets catalog carry a category badge linking to the workflow stage. Others open on Hugging Face.

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Metadata

License
apache-2.0
Tags
mlx safetensors sensenova_loop 4-bit nvfp4 abliterated arete sense-nova-loop text-generation conversational custom_code dataset:NousResearch/arete-data-pairs

Related

Total size
7.43 GB
Files
14
Quantizations
1
Registered
2026-11-18 12:04
Last updated on HF
2026-11-18 11:54

Files by quantization

Auxiliary files 14 files 7.45 GB
model.safetensors 7.43 GB dfd5b117 download
tokenizer.json 19.1 MB 06b95093 download
modeling_sensenova_loop.py 18.5 KB b437a10f download
chat_template.jinja 7.57 KB a585dec8 download
README.md 4.95 KB 710b0a29 download
load_nvfp4.py 4.23 KB 3f68b0b0 download
config.json 3.59 KB 3c54ecaf download
.gitattributes 1.53 KB 52373fe2 download
processor_config.json 1.16 KB 33818c7f download
tokenizer_config.json 1.14 KB 1d134cd2 download
video_preprocessor_config.json 550 B a79976a7 download
preprocessor_config.json 443 B 8ed39680 download
generation_config.json 198 B 01ef5fe7 download
mlx_convert.json 170 B 3fff19b3 download

README current version from Hugging Face


library_name: mlx
license: apache-2.0
pipeline_tag: text-generation
base_model: sensenova/SenseNova-Loop-9B
base_model_revision: 329a65f701ba8c48b73c30d0cbaa6f0449733673
datasets:

  • NousResearch/arete-data-pairs
  • HuggingFaceFW/fineweb-edu
    tags:
  • mlx
  • 4-bit
  • nvfp4
  • abliterated
  • arete
  • sense-nova-loop

SenseNova-Loop-9B abliterated + Arete, NVFP4 -> MLX 4-bit

The Apple Silicon build of
balllvin/SenseNova-Loop-9B-abliterated-arete-NVFP4:
the final NVFP4 build converted to MLX affine 4-bit (7.4 GB), with the looped 56-of-32 layer plan and the
mHC routers ported to MLX. It is the final MLX build of this family. It is for research.

Builds

Build Repo
QAD base (NVFP4) balllvin/SenseNova-Loop-9B-NVFP4
abliterated (NVFP4) balllvin/SenseNova-Loop-9B-abliterated-NVFP4
+ Arete (NVFP4) balllvin/SenseNova-Loop-9B-abliterated-arete-NVFP4
QAD base, MLX 4-bit balllvin/SenseNova-Loop-9B-NVFP4-MLX-4bit
+ Arete, MLX 4-bit this repo

Speed and memory

M5 Pro, 64 GB, 48-64 tokens out, load 7-15 from other work, so conservative:

Mode Steady footprint Decode tok/s End-to-end, 64 tok TTFT
default, 8-bit head 7.5 GB 20.8 18.0 0.44 s
--fast-head ~8.0 GB 26.6 22.8 0.25-0.32 s

The original MLX build of these weights measured 8.1 tok/s. With --fast-head and --max-batch 8,
server-path concurrency is 22.6 / 33.2 / 53.9 / 51.3 total tok/s at 1 / 2 / 4 / 8 streams (per stream 22.6 /
16.6 / 13.5 / 6.4-11.3). The server is on demand: it starts with the model, warms up the common shapes, and
exits after 30 minutes idle.

Local MLX gate

On 160 teacher-forced WikiText positions:

  • Router + bf16 trunk: NLL 3.04790 against 3.04762 for the original iterative fp32 path (+0.00028),
    top-1 98.1%.
  • Default 8-bit head: NLL 3.04775 against 3.04790 for bf16 (-0.00015), top-1 99.4%.
  • Fast head: NLL -0.00079 against bf16, top-1 199/199, mean KL 0.000544.

The sealed refusal/MMLU/NLL results are on the NVFP4 card;
they were not re-run on this MLX conversion.

Run it

python tools/serve.py --model . --port 8099 --kv-bits 8 --max-batch 4 --loops 2 --head-bits 8
# faster, +0.57 GB
python tools/serve.py --model . --port 8099 --kv-bits 8 --max-batch 8 --loops 2 --fast-head

The server renders the full message list, supports SSE streaming, and exits after 30 minutes idle
(--idle-exit 0 keeps it up). --loops 1 is faster (27.3 ms/token) but WikiText NLL is 3.921 against 3.041
at two loops; 2 is the trained setting.

How it was made

  1. Conversion: the Arete NVFP4 build to MLX affine 4-bit (group 64), 7.4 GB.
  2. Port: the looped execution plan (56 visits of 32 physical layers) and the mHC routers; the gates stay
    fp32 but return bf16, so the trunk runs bf16.
  3. Kernels: the checkpoint's 20 alternating 2x2 normalisations are a Mobius map; their exact k-step
    closed form is compiled for the decode shape, and the LM head is 8-bit by default (fast head optional).
  4. Server: on demand, with warmup, streaming and loop selection.

Limits

  • The sealed refusal/MMLU/NLL table is the NVFP4 source build; it was not re-run on the MLX conversion. The
    local gate above is 160 WikiText positions.
  • Abliteration is behavioural; it answers 93 of 99 harmful prompts in the sealed set. It is not
    safety-aligned.
  • --loops 1 is not the trained model.
  • Speed and memory are one M5 Pro; the load was 7-15 from other work, so the numbers are conservative.

Credits

balllvin/SenseNova-Loop-9B-abliterated-arete-NVFP4
for the source build; sensenova/SenseNova-Loop-9B for
the base; NousResearch/arete-data-pairs for
Arete; FineWeb-Edu for the QAD corpus;
MLX and mlx-lm for the runtime;
JailbreakBench and
AdvBench for the sealed prompts on the NVFP4 card.

Licence

The weights are a modified version of SenseNova-Loop-9B
under Apache-2.0. The Arete data states no licence; its makers
released it publicly. This is an independent community build, unaffiliated with the model's authors.

README history 13 versions

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-11-18 card 0760d51 6.6 KB
    Loading...
  2. 2026-11-18 card: state the build is text-only (vision tower not converted to MLX) 517caef 5.2 KB
    Loading...
  3. 2026-11-18 card: set base_model_relation per BrainSpa model-tree rule 576ce79 5 KB
    Loading...
  4. 2026-11-18 card: MLX repos renamed, -4bit suffix dropped 0767918 4.9 KB
    Loading...
  5. 2026-11-18 card: match the Flint release structure; cross-link the MLX and NVFP4 builds d4f54ef 4.9 KB
    Loading...
  6. 2026-11-18 card: low-RAM default and on-demand server 3256fa9 5.1 KB
    Loading...
  7. 2026-11-18 low-RAM default: 8-bit head, on-demand server, idle exit 99344a4 5 KB
    Loading...
  8. 2026-11-18 model-id @N loop selector; final tools 5a57792 4.6 KB
    Loading...
  9. 2026-11-18 bscode wrapper, LaunchAgent, final card 68a322c 4.4 KB
    Loading...
  10. 2026-11-18 warmup at startup; 26.6 tok/s decode 1aca634 3.9 KB
    Loading...
  11. 2026-11-18 26 tok/s: closed-form mHC router, bf16 trunk, two-stage head; loops option 163677e 3.7 KB
    Loading...
  12. 2026-11-18 final abliterated+Arete MLX 4-bit 7cbfed3 1.6 KB
    Loading...
  13. 2026-11-18 QAD+Arete NVFP4 -> MLX 4-bit 19a3b1e 1.1 KB
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration