Repair optimized quality vision compatibility

Replace the broken optimized-quality root trunk with the official mlx-community/Qwen3.6-27B-8bit VLM trunk, move the MTPLX draft sidecar to mtp/weights.safetensors, and remove stale root files that generic VLM loaders misread. Local mlx_vlm text/image/video checks passed before upload; MTPLX can load the nested sidecar under the existing conservative runtime contract.

README.md

@@ -31,7 +31,7 @@ This is intended as the higher-quality sibling to `Qwen3.6-27B-MTPLX-Optimized-S
 mtplx start --model Youssofal/Qwen3.6-27B-MTPLX-Optimized-Quality
 ```
 
-The artifact includes `mtplx_runtime.json` and `mtp.safetensors`, so MTPLX can inspect and route it through the native Qwen MTP backend.
+The artifact includes `mtplx_runtime.json` and `mtp/weights.safetensors`, so MTPLX can inspect and route it through the native Qwen MTP backend while generic MLX vision loaders only glob the base model shards.
 
 ## Local bakeoff snapshot
 
@@ -51,3 +51,7 @@ This row is a local release-readiness check, not a broad public hardware claim.
 - Base model: `Qwen/Qwen3.6-27B`
 - INT8 MTP source: `cyankiwi/Qwen3.6-27B-AWQ-BF16-INT8`
 - MTPLX staging manifest: `mtplx_upload_manifest.json`
+
+## Vision compatibility
+
+This revision keeps the optimized MTPLX draft sidecar at `mtp/weights.safetensors` so generic VLM loaders see only the normal Qwen vision/text weight shards at the repository root. The base trunk is the MLX 8-bit Qwen3.6 vision layout; MTPLX reads the draft sidecar through `config.json`.

model-00001-of-00006.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f3823b758b2963a0ad5c4ef54b1e09066eb95f8d1025ec2d0bf00591d73121e5
-size 5305469313
+oid sha256:f824099df8957f3d5e773c2faad5316ed6e43cdc8e41497e4f6c96bb87bb160d
+size 5317707349

model-00002-of-00006.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4ac0fcc806775859c0273f4ba4a8117335eabaaf2cc4e5d63092b8955ce4b0a0
-size 5354102628
+oid sha256:6b0ac8faafe5ed25f5c19ffe74b5ea9980c18d04baf4d98169db1c67a2ed35e5
+size 5354102528

model-00003-of-00006.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3deafdb3efed804fac4f3946991fb16ce1776dcab4611f7e6ec8375c94a4f41a
-size 5337392986
+oid sha256:4923f72bbfcfab5749a9a8d88f654515bbf15e3f6f01f3199ff67a5f50c58f85
+size 5354184654

model-00004-of-00006.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c99b0324ab5834d4fd8182d32e4250e1e50b199381e860a09cfcf8c5bc4a30df
-size 5292847128
+oid sha256:e0ee725dd0fc3cc7ca2a94734f18f9ab5ea3bdfa0c749353687d6f3442c0ab64
+size 5337309675

model-00005-of-00006.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:98a0f2aca3592183b50c92dc4581ee5f1a3288f5da4258a6dec6feee7142bbaa
-size 5354102635
+oid sha256:e6731a152f2ec3773526354b7104e9be1231b2bc9b1ec9cd033e2cc4470480ec
+size 5292848520

model-00006-of-00006.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fbe7b8f7819e5faacbfcca0da570285891a5ed1ba217c08c8b8b33fd50075ffa
-size 1935805825
+oid sha256:5c8550886d4bd9249079cf1cb1e4cd572e8e491f5db60ae8a78e957be1dadb24
+size 2845065535

model-vision-00001-of-00001.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:2be2da5032ebb3f56ccd3738ba1f93f78c61b5f29ae118197559b009a802cddd
-size 921497261

mtplx_upload_manifest.json

@@ -1,31 +1,24 @@
 {
-  "benchmark": {
-    "acceptance_by_depth": [
-      0.9556737588652481,
-      0.852575488454707,
-      0.7406749555950266
-    ],
-    "artifact": "/tmp/mtplx-flat8-maxfan-bakeoff-20260510-040405/flat8-int8-flappy2k.json",
-    "peak_gib": 27.615886019542813,
-    "suite": "flappy 2k depth3 max-fan",
-    "tok_s": 33.6276837588148,
-    "verify_ms_per_call": 88.13897531140259
-  },
-  "created_from": "/Users/youssof/Documents/MTPLX/models/Qwen3.6-27B-MTPLX-Flat8-Quality-TargetBase",
+  "base_repo": "mlx-community/Qwen3.6-27B-8bit",
+  "base_revision": "c5a593c1475a746e43a543b0a02bd2b357e5745f",
   "linked_files": [
+    ".gitattributes",
+    "README.md",
     "chat_template.jinja",
+    "config.json",
     "configuration.json",
     "generation_config.json",
-    "merges.txt",
     "model-00001-of-00006.safetensors",
     "model-00002-of-00006.safetensors",
     "model-00003-of-00006.safetensors",
     "model-00004-of-00006.safetensors",
     "model-00005-of-00006.safetensors",
     "model-00006-of-00006.safetensors",
-    "model-vision-00001-of-00001.safetensors",
     "model.safetensors.index.json",
+    "mtp/weights.safetensors",
     "mtplx_build_metadata.json",
+    "mtplx_runtime.json",
+    "mtplx_upload_manifest.json",
     "preprocessor_config.json",
     "processor_config.json",
     "tokenizer.json",
@@ -34,13 +27,26 @@
     "vocab.json"
   ],
   "mtp": {
-    "converted_sidecar": "/tmp/mtplx-flat8-maxfan-bakeoff-20260510-040405/cyankiwi-int8-mtp-g128.safetensors",
-    "policy": "INT8 affine group128, BF16 aux preserved",
+    "path": "mtp/weights.safetensors",
     "sha256": "dff5112ffdeff420d38f9227cc00f40dc302e009e70a3e8dc85c3c443b6364ab",
-    "source_file": "/tmp/hf-cyankiwi-qwen36-int8/model-00008-of-00008.safetensors",
-    "source_repo": "cyankiwi/Qwen3.6-27B-AWQ-BF16-INT8",
+    "size": 488790135,
+    "source": "previous optimized-quality mtp.safetensors",
     "tensor_count": 29
   },
+  "removed_remote_files": [
+    "mtp.safetensors",
+    "model-vision-00001-of-00001.safetensors",
+    "merges.txt"
+  ],
   "repo_id": "Youssofal/Qwen3.6-27B-MTPLX-Optimized-Quality",
-  "target": "flat8 affine group64"
+  "target": "flat8 affine group64 vision-compatible MLX trunk",
+  "validation": {
+    "mlx_vlm_image_passed": true,
+    "mlx_vlm_text_passed": true,
+    "mlx_vlm_video_passed": true,
+    "mtplx_contract_status": "runtime contract remains conservative; exactness_baseline is still candidate-promoted-by-user-decision",
+    "mtplx_depth": 1,
+    "mtplx_generation_mode": "mtp",
+    "mtplx_nested_sidecar_generation_text": "HELLO"
+  }
 }

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523
-size 19989325
+oid sha256:87a7830d63fcf43bf241c3c5242e96e62dd3fdc29224ca26fed8ea333db72de4
+size 19989343

tokenizer_config.json

@@ -10,7 +10,6 @@
   "errors": "replace",
   "image_token": "<|image_pad|>",
   "is_local": true,
-  "local_files_only": false,
   "model_max_length": 262144,
   "model_specific_special_tokens": {
     "audio_bos_token": "<|audio_start|>",
@@ -23,9 +22,9 @@
   },
   "pad_token": "<|endoftext|>",
   "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "processor_class": "Qwen3VLProcessor",
   "split_special_tokens": false,
-  "tokenizer_class": "Qwen2Tokenizer",
-  "tool_parser_type": "qwen3_coder",
+  "tokenizer_class": "TokenizersBackend",
   "unk_token": null,
   "video_token": "<|video_pad|>",
   "vision_bos_token": "<|vision_start|>",