← back to catalog · registered 2026-11-20 21:04

AtomicChat/Inkling-2-Mini-Uncensored-GGUF

AtomicChat GGUF multimodal
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/AtomicChat%2FInkling-2-Mini-Uncensored-GGUF"
Response includes
  • classification m-uncensored
  • files 5
  • author_summary 30 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M-U
Primary method

Uncensored (method unknown)

No other method signals detected in this model.
Confidence
LOW
Why this label 3 signals
Weak or ambiguous signals. Best guess based on catalog patterns; treat as tentative and check the evidence below.
  • 'uncensored' in name/tags but no 'abliterated' marker
  • method not identifiable from author declaration alone
  • may be DPO fine-tune, prompt engineering, or unknown technique
Refusal direction extraction

No specific extraction method could be identified for this model. The producer either did not document it or used a proprietary pipeline.

What is a refusal direction? →
Downloads · 30-day
10K
Likes
285
Descendants
2
in 2 direct forks
Model age
1d ago
created 2026-11-20
Downloads over time
Now 9.5K → from 0 ↑ 0%
03.5K7K10.5K 0 on Nov 20 9.5K on Nov 22 Nov
Nov 20 → Nov 22 · 3 snapshots · spans 2 days

Genealogy 2 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
gguf llama.cpp inkling2 uncensored lora image-text-to-text base_model:thinkingmachines/Inkling-2-Mini base_model:adapter:thinkingmachines/Inkling-2-Mini license:apache-2.0 endpoints_compatible region:us imatrix

Related

Total size
0 B
Files
5
Quantizations
3
Registered
2026-11-20 21:04
Last updated on HF
2026-11-21 17:57

Files by quantization

BF16 1 file 612 MB
mmproj-Inkling-2-Mini-Uncensored-BF16.gguf 612 MB 7dd7fd3a download
F16 1 file 610 MB
mmproj-Inkling-2-Mini-Uncensored-F16.gguf 610 MB 5a72e0b6 download
Auxiliary files 3 files 15.2 KB
LICENSE 11.1 KB d6456956 download
README.md 2.49 KB b9e3d53d download
.gitattributes 1.64 KB 449e6172 download

README current version from Hugging Face


license: apache-2.0
base_model: thinkingmachines/Inkling-2-Mini
pipeline_tag: image-text-to-text
tags:

  • gguf
  • llama.cpp
  • inkling2
  • uncensored
  • lora

Inkling-2-Mini Uncensored GGUF

Atomic Chat Discord GitHub X
  • Built by the team behind Atomic Chat, the app that runs open-weight models locally and offline.

An uncensored build of thinkingmachines/Inkling-2-Mini by AtomicChat. The refusal direction is removed from the weights that write into the residual stream. The base model is a 26.4B mixture-of-experts model (3.4B active) with a looped core. It takes text and image input. License: Apache-2.0, the same as the base model (LICENSE in this repo).

Status: in progress. The files will land here: a LoRA adapter for llama.cpp (--lora-scaled) and ready-made GGUF files with the edit already applied. This card will then report the measured results:

  • refusals before and after;

  • XSTest (false refusals on safe prompts);

  • MMLU;

  • KLD against the original model.

  • Runtime. The looped Inkling 2 needs llama.cpp with ggml-org/llama.cpp#32695, a draft PR at the time of writing. llama.cpp releases without that PR do not load this model.

  • Original quants: AtomicChat/Inkling-2-Mini-GGUF.

This model will answer requests that the original refuses. You are responsible for how you use it.

README history 14 versions

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

  1. 2026-11-21 Card: link the how-to-run guide on atomic.chat 40886ab 7.6 KB
    Loading...
  2. 2026-11-21 Update README.md 8da1bfa 7.3 KB
    Loading...
  3. 2026-11-21 Card: b12600+ does not load these files yet db6ed5a 7.3 KB
    Loading...
  4. 2026-11-21 Card: random-edit KLD for comparison 4d35d78 7.1 KB
    Loading...
  5. 2026-11-21 Card: every gate measured; MMLU and first-token KL misses disclosed 9924964 7 KB
    Loading...
  6. 2026-11-21 Card: MMLU (a gate miss, disclosed) 3f54e6f 6.4 KB
    Loading...
  7. 2026-11-21 Card: refusals with long reasoning, tool calls d65e323 6.1 KB
    Loading...
  8. 2026-11-21 Card: baked AD-IQ2_M f0f8b34 5.9 KB
    Loading...
  9. 2026-11-21 Card: test-set refusals, XSTest, baked AD-IQ3_S 2200d14 5.7 KB
    Loading...
  10. 2026-11-21 Card: baked AD-Q4_K 6e05c39 5.3 KB
    Loading...
  11. 2026-11-21 Card: validation refusals and first-token KL of the adapter 2377cd3 4.5 KB
    Loading...
  12. 2026-11-21 Card: early release, the LoRA adapter and what is measured so far 4eb44fe 3.9 KB
    Loading...
  13. 2026-11-20 card: Atomic Chat, Discord, GitHub and X buttons; Atomic Chat bullet 7f3577d 2.5 KB
    Loading...
  14. 2026-11-20 Card: status while the files are built 4d0325f 1.3 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