Downloads · 30 days
511
48% of all-time downloads
Blackfrost-AI/Qwen-3.8-2.4T-A95B-Derisked-IQ1_0-GGUF
Qwen-3.8-2.4T-A95B-Derisked-IQ1_0-GGUF is a text generation model from Blackfrost-AI. Use it when you need the model to write or continue text. It is set up for llama.cpp. The card lists the license as other.
Downloads · 30 days
511
48% of all-time downloads
All-time downloads
1.1K
Public
Repo size
397 GB
Likes
6
Public
Click a slice to open those files.
.gguf397 GB · 100%
From the Hugging Face model README
Qwen3.8-Max is a 2.4T-parameter mixture of experts. In BF16 it is a multi-terabyte, multi-node deploy, and its stock refusal behaviour is tuned for a general consumer assistant — which means it declines work that specialist teams do legitimately every day. Security engineers hunting bugs, red teams, and domain researchers hit refusals on tasks that are squarely inside their remit.
This checkpoint addresses both at once:
Footprint — importance-matrix-guided GGUF at ~397 GB, the smallest edition Blackfrost
publishes of this model. It brings a 2.4T MoE onto workstation-class GPUs rather than a
data-centre fabric deploy, and runs under llama.cpp rather than a multi-node serving stack.
Behaviour — Directional Weight Modification (DWM) applied to soften over-refusal on legitimate domain work, without removing the model's embedded safety protocols.
Safety behaviour can be further adjusted by domain specific chat templates.
The point is a model that stays useful inside a domain and stays safe. See Safety posture — that section is not boilerplate, it is the design.
Choosing between the GGUF editions: this one targets RTX-class hardware. The sibling
…-DERISKED-UD-IQ3_XXS carries more bits per weight and targets a B200-class
node. Pick by the hardware you have.
| Architecture | Qwen3_5MoeForCausalLM (qwen3_5_moe_text) — MoE + hybrid/linear attention |
| Base | Qwen/Qwen3.8-2.4T-A95B — official |
| Parameters | ~2.4T total · ~95B active |
| Layers / experts | 92 layers · 512 routed experts |
| Transform | DWM (directional weight modification) + imatrix GGUF quantization |
| Precision | IQ1_0, importance-matrix guided · mixed per-tensor allocation |
| Kept at higher precision | attention output · shared experts · normalisation · embeddings |
| On-disk | ~397 GB (370 GiB), single GGUF |
| Context | 262,144 native |
| Runtime | llama.cpp / llama-server |
| Serve shape | single node — 8× RTX PRO 6000 Blackwell (96 GB) recommended |
Not an ablation, and not an uncensored model.
Blackfrost calls this process DWM — Directional Weight Modification. Traditional abliteration attempts to delete refusal wholesale. DWM does something narrower and reversible in principle: it identifies a behavioural direction in the model's representation space and softens the model's response along that one axis, by a controlled amount, on selected residual-write surfaces.
What that buys, in practice:
No SFT, no DPO, no distillation, and no re-training of any kind was applied.
Transform parameters are not published in this card. The specific direction set, coefficient, and pass structure are Blackfrost method and are withheld. What is disclosed is what was changed (surfaces above) and what was not (everything else).
Read this before evaluating.
This is a de-risked model. Refusal behaviour has been deliberately modified at the weight level. If your approval process treats reduced-refusal models as a distinct category, this one belongs in that category — unlike Blackfrost's pruned-only releases, which carry their parent's safety behaviour unmodified.
What that does not mean:
Safety is designed as two layers, and the second layer is your responsibility:
Serving this model with an empty or generic system prompt discards half the design. If you are exposing it to customers or employees, layer 2 is not optional.
| Base | Official Qwen/Qwen3.8-2.4T-A95B |
| Applied | DWM (directional weight modification) · importance-matrix GGUF quantization |
| Quantized from | the full-precision BF16 build — not transcoded from a lower-precision edition |
| Not applied | SFT · DPO · distillation · expert pruning (REAP) · router modification |
| Format | GGUF · IQ1_0 · single file |
| Sibling editions | …-DERISKED-UD-IQ3_XXS (B200-class GGUF) · …-DERISKED-W4A4-NVFP4 · …-DERISKED-FP8 · …-DERISKED-BF16 |
[PENDING — not yet published.]
| Benchmark | DERISKED-BF16 | This (DERISKED-IQ1_0) | Retention |
|---|---|---|---|
| — | pending | pending | —% |
Upstream Qwen3.8 figures are deliberately not reproduced here. Quoting a parent model's benchmarks on a derivative card tells the reader nothing about this checkpoint. Blackfrost publishes its own measurements against its own harness, or it publishes nothing.
Both quantization and DWM are trades, and this is the most aggressively compressed edition Blackfrost publishes of this model — the compression trade here is larger than on the higher-precision editions. When results land, this section will state the harness, the conditions, and the retention figures — including any benchmark where retention is poor, and including multilingual capability, which generic English-only evaluations do not capture. Retention is measured against the BF16 edition, so the quantization cost is reported separately from the DWM effect.
--n-gpu-layers.llama-server --model at that file.max_tokens or visible content can come back empty while the model is still reasoning.llama-server \
--model "$MODEL_DIR/Qwen-3.8-2.4T-A95B-Derisked-IQ1_0.gguf" \
--ctx-size 32768 \
--n-gpu-layers 999 \
--host 0.0.0.0 --port 8000
OpenAI-compatible: POST /v1/chat/completions, GET /v1/models.
Remember to supply your domain system prompt — see Safety posture.
This repository is manually gated. Access requests are reviewed by a person. Access is free; no purchase is required.
Ask us about DWM strength tuned to your workload, domain-specific system-prompt packages, or evaluation against your own harness rather than generic benchmarks.