SuperQwen3.8-Flash-Next-abliterated-GGUF

This is the measured GGUF deployment variant of Jiunsong/SuperQwen3.8-Flash-Next-abliterated@e7b3664cd9a5096a674db80f6ba3d379401cbc71. The derivative originates from Qwen/Qwen3.8-Flash-Next. It does not introduce a separate behavioral fine-tune; it converts the sealed SuperQwen payload using the immutable converter revision recorded below.

Variant

Format: GGUF
Quantization: {"fallback_tensor_count": 195, "mmproj": "Q8_0 with 27 F16 shape fallbacks", "model_bits_per_weight": 5.39, "model_tensor_count": 1224, "type": "Q4_K_M"}
Converter: ggerganov/llama.cpp@d7bd3bfcad3e29c7e49fd26f38c79ee3e9a3fd6b
Payload files: 9; payload bytes: 119754585760

Verified runtime

One real served inference call and the capability, tool-call, strict structured-output, reasoning, and formatting gates passed on Apple M5 Max, 128 GB unified memory, macOS using llama.cpp.

C1 decode: 44.234 tok/s
C1 prefill: 399.162 tok/s
Directly tested context: 4143 tokens
Recipe: llama-server -m ./SuperQwen3.8-Flash-Next-abliterated-Q4_K_M-00001-of-00008.gguf --port 8080

These numbers apply only to the sealed hardware, runtime, command, and artifact in release_evidence/runtime.json.

Integrity

The conversion and served runtime reports both bind to payload identity e18978cd43e357603d1a7d0b5ebd497a5a82fbc11482c4e04c778f87d8c02f02. SHA256SUMS.json seals every published file. Private-revision, anonymous public, and immutable-tag verification is recorded separately by the publisher.

Limitations

Direct context validation is bounded to 4,143 prompt tokens; the 262,144-token training declaration is not claimed as served.
The Q4_K_M output is 5.39 BPW; llama.cpp required documented mixed-precision fallback quantization for 195 of 1,224 tensors.
The Q8_0 multimodal projector uses F16 fallback for 27 shape-constrained tensors.
Native MTP and external speculative draft decoding were not served in this GGUF release and are not advertised.

This structural derivative changes refusal behavior relative to the upstream model. No compatibility, context, speed, quality, or universal behavior claim is made beyond the included measurements. Review the upstream Qwen license and terms before use.

Attribution

Upstream: Qwen/Qwen3.8-Flash-Next
Behavioral derivative: Jiunsong/SuperQwen3.8-Flash-Next-abliterated@e7b3664cd9a5096a674db80f6ba3d379401cbc71

SuperQwen3.8-Flash-Next-abliterated-GGUF

Variant

Format: GGUF

Quantization:

{"fallback_tensor_count": 195, "mmproj": "Q8_0 with 27 F16 shape fallbacks", "model_bits_per_weight": 5.39, "model_tensor_count": 1224, "type": "Q4_K_M"}

Converter: ggerganov/llama.cpp@d7bd3bfcad3e29c7e49fd26f38c79ee3e9a3fd6b

Payload files: 9; payload bytes: 119754585760

Verified runtime

One real served inference call and the capability, tool-call, strict structured-output, reasoning, and formatting gates passed on Apple M5 Max, 128 GB unified memory, macOS using llama.cpp.

C1 decode: 44.234 tok/s

C1 prefill: 399.162 tok/s

Directly tested context: 4143 tokens

Recipe: llama-server -m ./SuperQwen3.8-Flash-Next-abliterated-Q4_K_M-00001-of-00008.gguf --port 8080

These numbers apply only to the sealed hardware, runtime, command, and artifact in release_evidence/runtime.json.

Limitations

Direct context validation is bounded to 4,143 prompt tokens; the 262,144-token training declaration is not claimed as served.

The Q4_K_M output is 5.39 BPW; llama.cpp required documented mixed-precision fallback quantization for 195 of 1,224 tensors.

The Q8_0 multimodal projector uses F16 fallback for 27 shape-constrained tensors.

Native MTP and external speculative draft decoding were not served in this GGUF release and are not advertised.