d3-edge v3.1.0

MODEL_MANIFEST.json

@@ -7,11 +7,11 @@
  "attention_mode": "noncausal_full_attention",
  "max_input_tokens": null,
  "identity": {
-  "model_sha256": "5afbc0251ed54b60fa9ca704c4fe3141f3af5b8b3a0d0d1810785ecf13a3fdc2",
+  "model_sha256": "6e00d04abbdbdf6d35c9e12065769d9edaacc3bfe9404583891797718d239dc8",
   "files_sha256": {
    "chat_template.jinja": "273d8e0e683b885071fb17e08d71e5f2a5ddfb5309756181681de4f5a1822d80",
-   "model-00001-of-00001.safetensors": "602da669692760b1bb8d36ef6a0ba928cbf2cc1d3277a11663498726c6602358",
-   "readout.safetensors": "27d73d80a6890e4ed854a0933112239517f6ae079c24c8cc295acf77004fff18",
+   "model-00001-of-00001.safetensors": "91bb8618450a7f628f1432bbced76bc19ee7346c93d191a228fdd33c25a57a9d",
+   "readout.safetensors": "de33d5dc795bbd77168e40bfadc340cca9b9f163b88744c10e1a788f4e166d3d",
    "tokenizer.json": "5f9e4d4901a92b997e463c1f46055088b6cca5ca61a6522d1b9f64c4bb81cb42",
    "tokenizer_config.json": "49e2b6e395f959f077f1e992b338919c0d4a9732fc6e613995e06557f843500c"
   },
@@ -543,10 +543,10 @@
   }
  },
  "parameters": {
-  "text": 558394784,
+  "text": 493728704,
   "vision": 100592896,
   "readout": 261120,
-  "loaded": 659248800
+  "loaded": 594582720
  },
  "base_model": {
   "repo_id": "Qwen/Qwen3.5-0.8B",
@@ -578,18 +578,18 @@
    }
   ]
  },
- "built_utc": "2026-10-11T03:45:47+00:00",
+ "built_utc": "2026-10-11T05:44:01+00:00",
  "files_sha256": {
   "LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
   "NOTICE": "0e7694dee1927d83b3637be7a8f989b941ebd884dd40767b995c7adee484789b",
-  "README.md": "32854912de28825c9333ca9ba8287f666ae72a0180215bf28bd97a9e0e570f98",
+  "README.md": "ec1ef3df472b36b1392b9dd7fdc77119e85e86743e15b7818c29c18e1995b336",
   "assets/banner.png": "352353b5ac097e31e6447ac0db8185439894e9e6c66376565e30f33c39d27913",
   "assets/example-receipt.png": "b9dbb8b103bacd45f1de824d844a3c07b81d57cb1a7466712ef060ad7dcb3926",
   "assets/example-video.mp4": "f4968ef89d07e97f20df4fd8047510a7a8e9df5c2d07c35e363a97927f164ab9",
-  "assets/index-areas.png": "9e8471ba167fe866d286b9873dfa1f48d0a509adb507d11bf8d4242f9a35397c",
-  "assets/index-pareto.png": "90a4f22032d1f03caf8890357c7f182a92ffa44e33cc2b445b372bc129bed754",
+  "assets/index-areas.png": "06be306799c515e90bea74994788239a53c01c893183879b18192cebe0158531",
+  "assets/index-pareto.png": "13ee761164b7fe0e99a8a48a942d783a759196a4a1eb466b82ada6bc8ce36daa",
   "chat_template.jinja": "273d8e0e683b885071fb17e08d71e5f2a5ddfb5309756181681de4f5a1822d80",
-  "config.json": "c06c8fab4bbbd151934e82d8c835e85974d360d0782905a184b940461595ddee",
+  "config.json": "5e529d9aea2f8d387750acf5b02cc32a47d7e87dc65d6d008fba98e5d15acc4d",
   "d3_engine.py": "086238942762df45047a32cd9a8e2bedd9b63167eb40be260c776de567979f53",
   "d3_fast.py": "68a1fb9a2c3bde9ac2763bf504decae592ca12722698fbde1b39a31262955535",
   "d3_format.py": "187a0f359f6f89a844e723377769556f78133b118851e702e4b0fe5123ccb0f1",
@@ -598,12 +598,12 @@
   "d3_server.py": "f9b8f2481461a074f469b4f58da3b0d7004824096528366d5b1f23187dd6e692",
   "decision_config.json": "30b6095b945d5a93697868afdf9be9123ea892085fbaff3078112c48da20dc6b",
   "merges.txt": "a9d356d7bdf1ef4949e3e748e95b8e10ad9d4e2e838eddc38a0a7b6b94d1db8d",
-  "model-00001-of-00001.safetensors": "602da669692760b1bb8d36ef6a0ba928cbf2cc1d3277a11663498726c6602358",
-  "model.safetensors.index.json": "1b835c61450c3dde31d1ab6e8c187ee435dd7c6b152c571e9efe0d87a5f16aed",
+  "model-00001-of-00001.safetensors": "91bb8618450a7f628f1432bbced76bc19ee7346c93d191a228fdd33c25a57a9d",
+  "model.safetensors.index.json": "7d3951baba643a2c4fe2263d8602ef37b1c1f804b58949a5469449da6a588b03",
   "modeling_d3.py": "f8f8db23ff58f1a990ffa24977d744a50e1b4446d02a64c22070967705b6dbef",
   "pipeline_d3.py": "4f9297c7c74e596e9b67df91a1e6823cfada48bf781336f141974ff6a2f1038e",
   "preprocessor_config.json": "27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516",
-  "readout.safetensors": "27d73d80a6890e4ed854a0933112239517f6ae079c24c8cc295acf77004fff18",
+  "readout.safetensors": "de33d5dc795bbd77168e40bfadc340cca9b9f163b88744c10e1a788f4e166d3d",
   "requirements.txt": "57a1938448bc240bdf4fee5d1b9cbc8450109fa1f6cee11e94dc04f274be8a32",
   "tokenizer.json": "5f9e4d4901a92b997e463c1f46055088b6cca5ca61a6522d1b9f64c4bb81cb42",
   "tokenizer_config.json": "49e2b6e395f959f077f1e992b338919c0d4a9732fc6e613995e06557f843500c",
@@ -617,10 +617,10 @@
   "assets/banner.png": 2235551,
   "assets/example-receipt.png": 65918,
   "assets/example-video.mp4": 26300,
-  "assets/index-areas.png": 125963,
-  "assets/index-pareto.png": 186784,
+  "assets/index-areas.png": 123190,
+  "assets/index-pareto.png": 186971,
   "chat_template.jinja": 7755,
-  "config.json": 2650,
+  "config.json": 2572,
   "d3_engine.py": 4413,
   "d3_fast.py": 22331,
   "d3_format.py": 4881,
@@ -629,8 +629,8 @@
   "d3_server.py": 10100,
   "decision_config.json": 5504,
   "merges.txt": 3353259,
-  "model-00001-of-00001.safetensors": 1318015400,
-  "model.safetensors.index.json": 29351,
+  "model-00001-of-00001.safetensors": 1188677968,
+  "model.safetensors.index.json": 25448,
   "modeling_d3.py": 12079,
   "pipeline_d3.py": 2936,
   "preprocessor_config.json": 390,

README.md

@@ -22,18 +22,18 @@ tags:
 
 | | |
 | --- | --- |
-| **Parameters** | 0.66B, including the 0.10B vision encoder |
+| **Parameters** | 0.59B, including the 0.10B vision encoder |
 | **Inputs** | Text or JSON, plus images and videos (several per request) |
 | **Decision types** | Choice · Yes / No · Score |
 | **License** | Apache-2.0 |
 
 ## Highlights
 
-- **Jev Decision Index 0.3, public suite: 30.55**, measured with the official 0.3 kit on the released weights: all 140,178 public requests answered, none unsupported.
-- **+13.9 on the public suite over Decision 2.0** (its 0.6B model: 16.68 on the board), ahead in all five areas.
+- **Jev Decision Index 0.3, public suite: 28.82**, measured with the official 0.3 kit on the released weights: all 140,178 public requests answered, none unsupported.
+- **+12.1 on the public suite over Decision 2.0** (its 0.6B model: 16.68 on the board), ahead in all five areas.
 - **Reads images:** multiple images per request (PNG, JPEG or WebP), given as paths, URLs, PIL images or base64 data URLs; every question of the request sees all of them.
-- **Reads videos:** multiple videos per request (MP4, WebM, MOV or MKV), given as paths, URLs, base64 data URLs or frame arrays, read at 2 frames per second; **Perception Test (multiple-choice video QA): 47.5** (internal evaluation).
-- **Speed:** a median of 8.6 ms for a text request, 42.5 ms for a request with an image and 311.8 ms for a request with a 10-second video, on one AMD Instinct MI325X GPU, one request at a time.
+- **Reads videos:** multiple videos per request (MP4, WebM, MOV or MKV), given as paths, URLs, base64 data URLs or frame arrays, read at 2 frames per second; **Perception Test (multiple-choice video QA): 46.6** (internal evaluation).
+- **Speed:** a median of 6.7 ms for a text request, 41.9 ms for a request with an image and 308.3 ms for a request with a 10-second video, on one AMD Instinct MI325X GPU, one request at a time.
 - **Many questions, one call:** Choice, Yes / No and Score questions about the same input are answered together, each from its own forward pass over the input, with a probability for every option.
 
 ## Quickstart
@@ -148,7 +148,7 @@ Images go before the text of the request, each read at up to 1.6 megapixels; vid
 
 | Model | Jev Decision Index ↑ | Public ↑ | Same-skill tests ↑ | New-domain tasks ↑ |
 | --- | ---: | ---: | ---: | ---: |
-| **d3-edge** | **19.7** | **30.6** | **25.4** | **9.4** |
+| **d3-edge** | **18.0** | **28.8** | **23.9** | **8.0** |
 | Bekko System One v0 400M | 15.2 | 22.5 | 21.6 | 7.2 |
 | Dinah-0 | 13.9 | 26.1 | 18.8 | 5.5 |
 | Decision 2.0 (0.6B) | 13.6 | 16.7 | 16.9 | 11.8 |
@@ -163,33 +163,33 @@ Images go before the text of the request, each read at up to 1.6 megapixels; vid
 
 | Model | Vision Index ↑ | Public ↑ | Private ↑ |
 | --- | ---: | ---: | ---: |
-| **d3-edge** | **38.8** | **41.0** | **36.6** |
+| **d3-edge** | **35.6** | **37.5** | **33.6** |
 
 d3-edge on the public vision benchmarks († approximate rebuild):
 
 | Benchmark | d3-edge |
 | --- | ---: |
-| CV-Bench | 58.6 |
-| BLINK | 34.1 |
-| RealWorldQA | 27.6 |
-| CharXiv † | 46.9 |
-| InfographicVQA † | 70.0 |
-| Mind2Web † | 36.8 |
-| Winoground | 38.8 |
-| KIE (CORD+FUNSD) † | 88.6 |
-| Moderation (Hateful Memes) | 12.5 |
-| R-Bench-M | 9.0 |
-| MMMU-Pro vision | 5.3 |
+| CV-Bench | 54.2 |
+| BLINK | 35.1 |
+| RealWorldQA | 24.8 |
+| CharXiv † | 40.9 |
+| InfographicVQA † | 60.7 |
+| Mind2Web † | 41.2 |
+| Winoground | 29.0 |
+| KIE (CORD+FUNSD) † | 86.4 |
+| Moderation (Hateful Memes) | 6.8 |
+| R-Bench-M | 10.3 |
+| MMMU-Pro vision | 4.9 |
 
 ### Videos: Perception Test
 
 | Perception Test (validation) | d3-edge |
 | --- | ---: |
-| **All 19,140 questions** | **47.5** |
-| Memory | 40.9 |
-| Abstraction | 37.8 |
-| Physics | 36.0 |
-| Semantics | 71.5 |
+| **All 19,140 questions** | **46.6** |
+| Memory | 40.6 |
+| Abstraction | 36.2 |
+| Physics | 38.3 |
+| Semantics | 67.3 |
 
 Multiple-choice video question answering (three options per question, chance 33.3): top-1 accuracy (%), videos read with the defaults above.
 

config.json

@@ -28,11 +28,8 @@
       "linear_attention",
       "linear_attention",
       "full_attention",
-      "linear_attention",
       "full_attention",
-      "linear_attention",
       "full_attention",
-      "linear_attention",
       "full_attention"
     ],
     "linear_conv_kernel_dim": 4,
@@ -47,7 +44,7 @@
     "mtp_num_hidden_layers": 1,
     "mtp_use_dedicated_embeddings": false,
     "num_attention_heads": 8,
-    "num_hidden_layers": 15,
+    "num_hidden_layers": 12,
     "num_key_value_heads": 2,
     "pad_token_id": null,
     "partial_rotary_factor": 0.25,

model-00001-of-00001.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:602da669692760b1bb8d36ef6a0ba928cbf2cc1d3277a11663498726c6602358
-size 1318015400
+oid sha256:91bb8618450a7f628f1432bbced76bc19ee7346c93d191a228fdd33c25a57a9d
+size 1188677968

model.safetensors.index.json

@@ -1,6 +1,6 @@
 {
   "metadata": {
-    "total_size": 1317975360
+    "total_size": 1188643200
   },
   "weight_map": {
     "language_model.embed_tokens.weight": "model-00001-of-00001.safetensors",
@@ -44,55 +44,16 @@
     "language_model.layers.10.self_attn.q_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.10.self_attn.v_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.11.input_layernorm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.11.linear_attn.A_log": "model-00001-of-00001.safetensors",
-    "language_model.layers.11.linear_attn.conv1d.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.11.linear_attn.dt_bias": "model-00001-of-00001.safetensors",
-    "language_model.layers.11.linear_attn.in_proj_a.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.11.linear_attn.in_proj_b.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.11.linear_attn.in_proj_qkv.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.11.linear_attn.in_proj_z.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.11.linear_attn.norm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.11.linear_attn.out_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.11.mlp.down_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.11.mlp.gate_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.11.mlp.up_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.11.post_attention_layernorm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.input_layernorm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.mlp.down_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.mlp.gate_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.mlp.up_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.post_attention_layernorm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.self_attn.k_norm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.self_attn.k_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.self_attn.o_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.self_attn.q_norm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.self_attn.q_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.12.self_attn.v_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.input_layernorm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.linear_attn.A_log": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.linear_attn.conv1d.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.linear_attn.dt_bias": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.linear_attn.in_proj_a.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.linear_attn.in_proj_b.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.linear_attn.in_proj_qkv.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.linear_attn.in_proj_z.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.linear_attn.norm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.linear_attn.out_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.mlp.down_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.mlp.gate_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.mlp.up_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.13.post_attention_layernorm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.input_layernorm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.mlp.down_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.mlp.gate_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.mlp.up_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.post_attention_layernorm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.self_attn.k_norm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.self_attn.k_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.self_attn.o_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.self_attn.q_norm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.self_attn.q_proj.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.14.self_attn.v_proj.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.11.self_attn.k_norm.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.11.self_attn.k_proj.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.11.self_attn.o_proj.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.11.self_attn.q_norm.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.11.self_attn.q_proj.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.11.self_attn.v_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.2.input_layernorm.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.2.mlp.down_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.2.mlp.gate_proj.weight": "model-00001-of-00001.safetensors",
@@ -183,19 +144,16 @@
     "language_model.layers.8.self_attn.q_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.8.self_attn.v_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.9.input_layernorm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.9.linear_attn.A_log": "model-00001-of-00001.safetensors",
-    "language_model.layers.9.linear_attn.conv1d.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.9.linear_attn.dt_bias": "model-00001-of-00001.safetensors",
-    "language_model.layers.9.linear_attn.in_proj_a.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.9.linear_attn.in_proj_b.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.9.linear_attn.in_proj_qkv.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.9.linear_attn.in_proj_z.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.9.linear_attn.norm.weight": "model-00001-of-00001.safetensors",
-    "language_model.layers.9.linear_attn.out_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.9.mlp.down_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.9.mlp.gate_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.9.mlp.up_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.layers.9.post_attention_layernorm.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.9.self_attn.k_norm.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.9.self_attn.k_proj.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.9.self_attn.o_proj.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.9.self_attn.q_norm.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.9.self_attn.q_proj.weight": "model-00001-of-00001.safetensors",
+    "language_model.layers.9.self_attn.v_proj.weight": "model-00001-of-00001.safetensors",
     "language_model.norm.weight": "model-00001-of-00001.safetensors",
     "visual.blocks.0.attn.proj.bias": "model-00001-of-00001.safetensors",
     "visual.blocks.0.attn.proj.weight": "model-00001-of-00001.safetensors",

readout.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:27d73d80a6890e4ed854a0933112239517f6ae079c24c8cc295acf77004fff18
+oid sha256:de33d5dc795bbd77168e40bfadc340cca9b9f163b88744c10e1a788f4e166d3d
 size 1044560