SuperQwen3.8-Flash-Next-abliterated
Architecture-aware Qwen4-Exp refusal-subspace edit with measured capability preservation.
This is the original-weight BF16 derivative pinned to
Qwen/Qwen3.8-Flash-Next@f5d08274bafd880402bd16f5e3e6c514136ec06c.
It is not a LoRA and does not require an inference-time adapter.
Architecture
| Component | Verified checkpoint structure |
|---|---|
| Model class | Qwen4ExpForConditionalGeneration |
| Text stack | 48 layers, hidden size 2,560, four gated residual streams |
| Sequence blocks | 36 Gated DeltaNet + 12 Qwen Sparse Attention layers |
| MoE | 512 routed experts per layer, top-10, plus one shared expert |
| PLE | Bigram/trigram per-layer embedding at layer 2; preserved |
| Vision | 27-layer encoder; preserved |
| MTP | Native one-layer MTP tensors; preserved, not used in the measured K=0 recipe |
| Chat | Thinking controls, reasoning-effort controls, tool calls, multimodal input |
The edit targets the measured 2,560-dimensional attention/DeltaNet output maps plus every routed/shared-expert MLP down projection. Vision, PLE, expert routers and gates, hyper-connection gates, normalization, and MTP are outside the target set.
Method
- Corpus: 842 canonical harmful/harmless pairs.
- Hidden space: four-stream hyper-connection read mix, width 2,560.
- Selected candidate:
amo_r1_l12_a200. - Layers:
[33, 35, 36, 37, 38, 40, 41, 42, 44, 45, 46, 47]. - Output paths:
['attention', 'mlp']. - Rank: 1.
- Projection strength: 2.0.
- Changed storage tensor count: 36 (6168 logical output projections, including fused routed experts); zero router, PLE, vision, normalization, residual-gate, or MTP targets.
Verified BF16 release gates
| Gate | Result |
|---|---|
| Patched storage tensors | 36 |
| Logical output projections | 6,168 |
| Protected tensor families | PASS |
| Full artifact tensor/hash verification | PASS |
| Local served-inference benchmark | Not claimed |
The sealed BF16 artifact is 337 GiB. It cannot fit as an original-weight served checkpoint across the available two 128 GiB unified-memory DGX Spark nodes, so no BF16 throughput, context, or behavioral equivalence claim is made. The separately published FP8 variant applies the same logical target map and has independent served inference and behavioral evidence; quantization means its measurements are not presented as BF16 measurements.
Artifact access and compatibility status
hf download Jiunsong/SuperQwen3.8-Flash-Next-abliterated \
--revision v1.0.0 \
--local-dir ./model
hf cache verify Jiunsong/SuperQwen3.8-Flash-Next-abliterated \
--revision v1.0.0 \
--local-dir ./model \
--fail-on-missing-files \
--fail-on-extra-files
The exact-commit BF16 file set, index, tokenizer assets, and all 131 safetensors headers
were reloaded successfully. A generic Transformers 5.12.1 AutoConfig reload was also
tested and rejected model_type: qwen4_exp. Upstream Transformers commit
fc5c5bde8e656dad91cbf34e61940d984b1c7b91 adds the Qwen4-Exp implementation, but a complete
337 GiB BF16 model instantiation with that commit was not run on the available hardware.
Accordingly, this card does not claim released-Transformers compatibility or provide an
unverified AutoModel recipe. Validate a Qwen4-Exp-aware runtime against the immutable
artifact before deployment.
Verification evidence
| Evidence | SHA-256 |
|---|---|
| architecture report | 27b35581e97cc78f0a6e67f565aabc90f36fdd2908faeed285d6eb0fd88040ad |
| 842-pair corpus | 218c67b48f495eb559a36966067fa5101db2350f688c71f4b9b0ec7aecd6de53 |
| capture manifest | 408fb389190e6536cdb38eb3c9a8f5322ae15307552d2fe8ae04250535708273 |
| subspace report | b6bce6fd56bdaef8b5d766725d0dd55f2d6472f16e31e8768e0dc6ac2260fed2 |
| bounded candidate bank | 708b6b28d22b71cb2eb37c90cf0246abca806d29d4d6786611a065c603e78561 |
| selected candidate | 14aa5b9072c32513a84579e6ff14699c49887473b4484ae273851160db159f28 |
| baseline benchmark | 2d39fa15726b9738167d30421511d28e477be4bb5f87cd63ca70e50a1a1252cb |
| baseline full refusal stress | 86a66d642601ebb29ad5f43d1f83ffb3b9a0b0aee4aae595ec20adace118e148 |
| baseline matched refusal sample | 673c7a070d49c27dac5fbc4e973113fedc1cc5f8a95b0c7182d2f99dd80b179d |
| candidate matched refusal sample | 59d1b994d7be474ee591cd1ff861b6542b1aa18fb85217000c9b5150d1a624e1 |
| BF16 patch verification | 941e762a29c3ed2ba1ca54099c8c5be126ac0f40b19ab2ca7aa078c2f5ac3911 |
| BF16 artifact verification | 639764f46ae8a47b8fc2e7c556e762283ef3fb36bfd4f100137f467af805915f |
| FP8 patch verification | 9581118ec2f75871a1243ca4da49f16d9019fe1a0c0a89257e2821be902588d6 |
| FP8 artifact verification | 8434b1c2cf18022c701a4cf8f541aafdbf2c15633a5e065a5a2f36157c5b79e7 |
| FP8 benchmark | 6c2e282c05fe51aefdf3e80141237b065e95d03c648c3d240e235a9209d8f63b |
| FP8 long-context retries | 3089c621d716eadfc6402535325b05e4945f0586ee2528f989d3070d4ca1e0b2 |
| FP8 full refusal stress | db206b81a5b88535c2f063d89684851cee4785f79794cb5f0e05c01469fd20de |
| FP8 refusal anomaly retries | afe6f96e96bfd72d2f23b0a21e2f31b16275e7a60132572203ca2898ff02209e |
| FP8 clean-reload inference | a58f31ea261fbea20432720e62dc6714622f4a94f4832822d63a679957ccc440 |
| FP8 runtime recipe | 4691af15670888e0bc56071eacf43d13ea3d64c943bb0b1d9eccfa91145073c7 |
| FP8 runtime overlay | 92c2aac1245403ca0ad3ebd1db68dac3b478cd3aabd9b18884cca5c860876aa4 |
| FP8 rank-0 serve script | d9751b487dc1c88bd7b709a4625ecd7a8fb9467ad9157042f965b9f0257f79d5 |
| FP8 rank-1 serve script | d9751b487dc1c88bd7b709a4625ecd7a8fb9467ad9157042f965b9f0257f79d5 |
Upstream BF16 revision: f5d08274bafd880402bd16f5e3e6c514136ec06c
Official FP8 revision: bcd9f01ddc9cff2316eb84281bebcd5b058bddce
OBLITERATUS revision: a5a1ffa5849b442cf188b3c03fd4de71ddf5bdcc
A compact, path-sanitized copy of the validated metrics and runtime identities is
included at verification/RELEASE_EVIDENCE.json.
Limitations
- Abliteration reduces a measured refusal direction; it does not guarantee correctness, harmlessness, or universal compliance.
- The capability and behavior suites are finite regression gates.
- Native MTP exists in the checkpoint, but the reported deployment measurements used K=0.
- GGUF and MLX variants are not published because Qwen4-Exp converter/runtime support was not independently verified for this release.
- Operators remain responsible for access controls, policy, and downstream safeguards.
License and attribution
This derivative follows the upstream Qwen3.8-Flash-Next license. Review the included
LICENSE and the upstream repository before redistribution or deployment.