This is the MTPLX 2.0.1 wrapper for the Tess-4-27B abliterated MXFP4 vision MLX artifact. It includes the standard MLX/VLM body, the BF16 vision tower, and a native MTP sidecar for MTPLX speculative decoding.
Use the paired standard MLX repo when you want stock MLX-VLM or LM Studio compatibility without MTPLX.
Paired standard repo: Shiftedx/Tess-4-27B-Abliterated-MXFP4-Vision-MLX
License: Apache-2.0
Chat format: Qwen chat template with <think> reasoning blocks
Contents
Runtime format: MLX plus MTPLX metadata
Quantization: MXFP4, 4-bit, group size 32
Vision tower: BF16 vision tensors grafted from the source model
Vision tensor count: 333
MTP sidecar: mtp.safetensors
MTP tensor count: 15
MTPLX version used for validation: 2.0.1
MTPLX Validation
mtplx inspect --require-mtp passed locally after contract cleanup. The artifact is marked as native MTPLX-compatible.
Tune summary on Apple M4 Max, 64 GB unified memory, fans on auto:
Mode
Decode tok/s
Multiplier vs AR
Acceptance
AR
28.09
1.00x
n/a
D1
46.28
1.65x
0.979
D2
50.68
1.80x
0.970, 0.954
D3
42.20
1.50x
0.869, 0.717, 0.593
Recommended MTPLX depth: 2.
Abliteration Notes
The selected candidate used residual-direction weight orthogonalization against a Tess-specific refusal direction.
Setting
Value
Strength
2.5
Targets
attention, dense_down
Edited modules
128
Direction scope
global
Preserve column norm
true
Heldout screen, no code execution:
Variant
Refusal rate
Benign refusal rate
Utility pass rate
Avg generation tok/s
Parent MXFP4
1.00
0.00
1.00
23.91
Selected strength 2.5
0.00
0.00
1.00
24.22
Strength 3.0 trial
0.00
0.00
0.50
26.20
Strength 2.5 was selected because the 3.0 trial harmed utility in the heldout screen.
Vision Validation
mlx_vlm.generate smoke passed locally after the vision graft. The smoke image was described as:
A close-up of a white ceramic mug with a black handle, filled with dark coffee and topped with a swirl of foam.
BenchLocal Light Screen
Pack
Pass / Total
Score
Failed IDs
toolcall-15
11/15
73%
TC-03, TC-07, TC-11, TC-12
instructfollow-15
14/15
93%
IF-14
Total
25/30
83%
Important caveat: this was a non-canonical quick run with thinking disabled, max_tokens=2048, and timeout_per_case=90. It is a fast quality screen, not directly comparable to the upstream model-card full BenchLocal score of 122/150 for Tess-4-27B Q8.
