Added Paper link and Update scores: 28 cell corrections across 17 models
Browse files- app.py +2 -0
- data/results.json +231 -231
app.py
CHANGED
|
@@ -65,6 +65,8 @@ HEADER_HTML = """
|
|
| 65 |
<a href="https://github.com/Clemson-Capstone/VANTAGE-Bench" target="_blank" rel="noopener">GitHub</a>
|
| 66 |
<span class="lb-sep">|</span>
|
| 67 |
<a href="https://huggingface.co/datasets/nvidia/PhysicalAI-VANTAGE-Bench" target="_blank" rel="noopener">Dataset</a>
|
|
|
|
|
|
|
| 68 |
</nav>
|
| 69 |
<div class="lb-stats">
|
| 70 |
<span class="lb-stat"><b>3</b> Domains</span>
|
|
|
|
| 65 |
<a href="https://github.com/Clemson-Capstone/VANTAGE-Bench" target="_blank" rel="noopener">GitHub</a>
|
| 66 |
<span class="lb-sep">|</span>
|
| 67 |
<a href="https://huggingface.co/datasets/nvidia/PhysicalAI-VANTAGE-Bench" target="_blank" rel="noopener">Dataset</a>
|
| 68 |
+
<span class="lb-sep">|</span>
|
| 69 |
+
<a href="https://arxiv.org/pdf/2609.09396" target="_blank" rel="noopener">Paper</a>
|
| 70 |
</nav>
|
| 71 |
<div class="lb-stats">
|
| 72 |
<span class="lb-stat"><b>3</b> Domains</span>
|
data/results.json
CHANGED
|
@@ -1,8 +1,86 @@
|
|
| 1 |
{
|
| 2 |
"schema_version": "1.0",
|
| 3 |
"benchmark_version": "1.0",
|
| 4 |
-
"updated": "2026-
|
| 5 |
"models": [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
{
|
| 7 |
"id": "google-gemini-3-1-pro",
|
| 8 |
"name": "Gemini 3.1 Pro",
|
|
@@ -15,19 +93,45 @@
|
|
| 15 |
"is_new": false,
|
| 16 |
"date_evaluated": "2026-05-27",
|
| 17 |
"model_url": "https://deepmind.google/technologies/gemini/",
|
| 18 |
-
"overall":
|
| 19 |
-
"spatial":
|
| 20 |
-
"spatio_temporal":
|
| 21 |
-
"temporal": 40.
|
| 22 |
-
"semantic": 70.
|
| 23 |
-
"2d_localization":
|
| 24 |
"2d_referring_expressions": 67.17,
|
| 25 |
-
"2d_spatial_pointing":
|
| 26 |
-
"single_object_tracking":
|
| 27 |
"temporal_localization": 45.69,
|
| 28 |
-
"dense_video_captioning": 35.
|
| 29 |
-
"event_verification":
|
| 30 |
-
"video_qa": 71.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
},
|
| 32 |
{
|
| 33 |
"id": "alibaba-qwen3-vl-32b",
|
|
@@ -41,16 +145,16 @@
|
|
| 41 |
"is_new": false,
|
| 42 |
"date_evaluated": "2026-05-27",
|
| 43 |
"model_url": "https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct",
|
| 44 |
-
"overall": 55.
|
| 45 |
-
"spatial":
|
| 46 |
"spatio_temporal": 44.19,
|
| 47 |
-
"temporal":
|
| 48 |
"semantic": 65.67,
|
| 49 |
-
"2d_localization":
|
| 50 |
"2d_referring_expressions": 72.39,
|
| 51 |
"2d_spatial_pointing": 75.62,
|
| 52 |
"single_object_tracking": 44.19,
|
| 53 |
-
"temporal_localization":
|
| 54 |
"dense_video_captioning": 29.4,
|
| 55 |
"event_verification": 60.04,
|
| 56 |
"video_qa": 71.3
|
|
@@ -67,13 +171,13 @@
|
|
| 67 |
"is_new": false,
|
| 68 |
"date_evaluated": "2026-05-27",
|
| 69 |
"model_url": "https://huggingface.co/nvidia/Cosmos-Reason2-8B",
|
| 70 |
-
"overall": 54.
|
| 71 |
-
"spatial":
|
| 72 |
"spatio_temporal": 37.69,
|
| 73 |
"temporal": 39.9,
|
| 74 |
"semantic": 66.02,
|
| 75 |
"2d_localization": 83.88,
|
| 76 |
-
"2d_referring_expressions":
|
| 77 |
"2d_spatial_pointing": 68.6,
|
| 78 |
"single_object_tracking": 37.69,
|
| 79 |
"temporal_localization": 47.3,
|
|
@@ -81,6 +185,58 @@
|
|
| 81 |
"event_verification": 64.09,
|
| 82 |
"video_qa": 67.95
|
| 83 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 84 |
{
|
| 85 |
"id": "alibaba-qwen3-vl-8b",
|
| 86 |
"name": "Qwen3-VL-8B-Instruct",
|
|
@@ -93,11 +249,11 @@
|
|
| 93 |
"is_new": false,
|
| 94 |
"date_evaluated": "2026-05-27",
|
| 95 |
"model_url": "https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct",
|
| 96 |
-
"overall": 50.
|
| 97 |
"spatial": 67.26,
|
| 98 |
"spatio_temporal": 33.12,
|
| 99 |
"temporal": 36.98,
|
| 100 |
-
"semantic": 62.
|
| 101 |
"2d_localization": 59.93,
|
| 102 |
"2d_referring_expressions": 73.29,
|
| 103 |
"2d_spatial_pointing": 68.56,
|
|
@@ -119,7 +275,7 @@
|
|
| 119 |
"is_new": false,
|
| 120 |
"date_evaluated": "2026-05-27",
|
| 121 |
"model_url": "https://huggingface.co/Qwen/Qwen3.5-VL-27B-Instruct",
|
| 122 |
-
"overall": 49.
|
| 123 |
"spatial": 79.04,
|
| 124 |
"spatio_temporal": 24.4,
|
| 125 |
"temporal": 31.93,
|
|
@@ -145,46 +301,20 @@
|
|
| 145 |
"is_new": false,
|
| 146 |
"date_evaluated": "2026-05-27",
|
| 147 |
"model_url": "https://huggingface.co/Qwen/Qwen3.5-VL-9B-Instruct",
|
| 148 |
-
"overall": 48.
|
| 149 |
"spatial": 75.26,
|
| 150 |
"spatio_temporal": 23.52,
|
| 151 |
-
"temporal": 35.
|
| 152 |
-
"semantic": 59.
|
| 153 |
"2d_localization": 76.87,
|
| 154 |
"2d_referring_expressions": 77.16,
|
| 155 |
"2d_spatial_pointing": 71.74,
|
| 156 |
"single_object_tracking": 23.52,
|
| 157 |
"temporal_localization": 42.53,
|
| 158 |
-
"dense_video_captioning":
|
| 159 |
"event_verification": 46.78,
|
| 160 |
"video_qa": 71.63
|
| 161 |
},
|
| 162 |
-
{
|
| 163 |
-
"id": "google-gemini-3-2-flash-lite",
|
| 164 |
-
"name": "Gemini 3.1 Flash Lite",
|
| 165 |
-
"organization": "Google",
|
| 166 |
-
"params": "—",
|
| 167 |
-
"type": "closed",
|
| 168 |
-
"result_type": "single",
|
| 169 |
-
"verified": true,
|
| 170 |
-
"ensemble": false,
|
| 171 |
-
"is_new": false,
|
| 172 |
-
"date_evaluated": "2026-05-27",
|
| 173 |
-
"model_url": "https://deepmind.google/technologies/gemini/",
|
| 174 |
-
"overall": 51.335,
|
| 175 |
-
"spatial": 62.06,
|
| 176 |
-
"spatio_temporal": 46.68,
|
| 177 |
-
"temporal": 35.17,
|
| 178 |
-
"semantic": 61.43,
|
| 179 |
-
"2d_localization": 69.88,
|
| 180 |
-
"2d_referring_expressions": 52.62,
|
| 181 |
-
"2d_spatial_pointing": 63.68,
|
| 182 |
-
"single_object_tracking": 46.68,
|
| 183 |
-
"temporal_localization": 37.72,
|
| 184 |
-
"dense_video_captioning": 32.62,
|
| 185 |
-
"event_verification": 54.83,
|
| 186 |
-
"video_qa": 68.03
|
| 187 |
-
},
|
| 188 |
{
|
| 189 |
"id": "nvidia-cosmos-reason2-32b",
|
| 190 |
"name": "Cosmos-Reason2-32B",
|
|
@@ -197,12 +327,12 @@
|
|
| 197 |
"is_new": false,
|
| 198 |
"date_evaluated": "2026-05-27",
|
| 199 |
"model_url": "https://huggingface.co/nvidia/Cosmos-Reason2-32B",
|
| 200 |
-
"overall":
|
| 201 |
-
"spatial":
|
| 202 |
"spatio_temporal": 33.12,
|
| 203 |
-
"temporal": 40.
|
| 204 |
"semantic": 71.94,
|
| 205 |
-
"2d_localization":
|
| 206 |
"2d_referring_expressions": 50.04,
|
| 207 |
"2d_spatial_pointing": 74.03,
|
| 208 |
"single_object_tracking": 33.12,
|
|
@@ -212,82 +342,30 @@
|
|
| 212 |
"video_qa": 70.29
|
| 213 |
},
|
| 214 |
{
|
| 215 |
-
"id": "nvidia-cosmos3-
|
| 216 |
-
"name": "Cosmos3-
|
| 217 |
-
"organization": "NVIDIA",
|
| 218 |
-
"params": "64B",
|
| 219 |
-
"type": "open",
|
| 220 |
-
"result_type": "single",
|
| 221 |
-
"verified": true,
|
| 222 |
-
"ensemble": false,
|
| 223 |
-
"is_new": true,
|
| 224 |
-
"date_evaluated": "2026-05-29",
|
| 225 |
-
"model_url": "https://huggingface.co/nvidia/Cosmos3-Super",
|
| 226 |
-
"overall": 63.01333333,
|
| 227 |
-
"spatial": 78.71,
|
| 228 |
-
"spatio_temporal": 62.25,
|
| 229 |
-
"temporal": 40.72,
|
| 230 |
-
"semantic": 70.37,
|
| 231 |
-
"2d_localization": 86.97,
|
| 232 |
-
"2d_referring_expressions": 76.23,
|
| 233 |
-
"2d_spatial_pointing": 72.94,
|
| 234 |
-
"single_object_tracking": 62.25,
|
| 235 |
-
"temporal_localization": 51.9,
|
| 236 |
-
"dense_video_captioning": 29.54,
|
| 237 |
-
"event_verification": 71.28,
|
| 238 |
-
"video_qa": 69.46
|
| 239 |
-
},
|
| 240 |
-
{
|
| 241 |
-
"id": "nvidia-cosmos3-nano",
|
| 242 |
-
"name": "Cosmos3-Nano",
|
| 243 |
"organization": "NVIDIA",
|
| 244 |
-
"params": "
|
| 245 |
-
"type": "open",
|
| 246 |
-
"result_type": "single",
|
| 247 |
-
"verified": true,
|
| 248 |
-
"ensemble": false,
|
| 249 |
-
"is_new": true,
|
| 250 |
-
"date_evaluated": "2026-05-29",
|
| 251 |
-
"model_url": "https://huggingface.co/nvidia/Cosmos3-Nano",
|
| 252 |
-
"overall": 60.67291667,
|
| 253 |
-
"spatial": 74.84,
|
| 254 |
-
"spatio_temporal": 59.21,
|
| 255 |
-
"temporal": 39.73,
|
| 256 |
-
"semantic": 68.92,
|
| 257 |
-
"2d_localization": 74.11,
|
| 258 |
-
"2d_referring_expressions": 75.57,
|
| 259 |
-
"2d_spatial_pointing": 74.83,
|
| 260 |
-
"single_object_tracking": 59.21,
|
| 261 |
-
"temporal_localization": 48.04,
|
| 262 |
-
"dense_video_captioning": 31.42,
|
| 263 |
-
"event_verification": 68.88,
|
| 264 |
-
"video_qa": 68.95
|
| 265 |
-
},
|
| 266 |
-
{
|
| 267 |
-
"id": "alibaba-qwen3-vl-2b",
|
| 268 |
-
"name": "Qwen3-VL-2B-Instruct",
|
| 269 |
-
"organization": "Alibaba",
|
| 270 |
-
"params": "2B",
|
| 271 |
"type": "open",
|
| 272 |
"result_type": "single",
|
| 273 |
"verified": true,
|
| 274 |
"ensemble": false,
|
| 275 |
"is_new": true,
|
| 276 |
"date_evaluated": "2026-07-16",
|
| 277 |
-
"model_url": "https://huggingface.co/
|
| 278 |
-
"overall":
|
| 279 |
-
"spatial":
|
| 280 |
-
"spatio_temporal":
|
| 281 |
-
"temporal":
|
| 282 |
-
"semantic":
|
| 283 |
-
"2d_localization":
|
| 284 |
-
"2d_referring_expressions":
|
| 285 |
-
"2d_spatial_pointing":
|
| 286 |
-
"single_object_tracking":
|
| 287 |
-
"temporal_localization":
|
| 288 |
-
"dense_video_captioning":
|
| 289 |
-
"event_verification":
|
| 290 |
-
"video_qa":
|
| 291 |
},
|
| 292 |
{
|
| 293 |
"id": "nvidia-cosmos-reason2-2b",
|
|
@@ -301,12 +379,12 @@
|
|
| 301 |
"is_new": true,
|
| 302 |
"date_evaluated": "2026-07-16",
|
| 303 |
"model_url": "https://huggingface.co/nvidia/Cosmos-Reason2-2B",
|
| 304 |
-
"overall": 45.
|
| 305 |
-
"spatial":
|
| 306 |
"spatio_temporal": 26.85,
|
| 307 |
-
"temporal": 33.
|
| 308 |
"semantic": 59.98,
|
| 309 |
-
"2d_localization":
|
| 310 |
"2d_referring_expressions": 56.33,
|
| 311 |
"2d_spatial_pointing": 59.7,
|
| 312 |
"single_object_tracking": 26.85,
|
|
@@ -316,30 +394,30 @@
|
|
| 316 |
"video_qa": 64.69
|
| 317 |
},
|
| 318 |
{
|
| 319 |
-
"id": "
|
| 320 |
-
"name": "
|
| 321 |
-
"organization": "
|
| 322 |
-
"params": "
|
| 323 |
"type": "open",
|
| 324 |
"result_type": "single",
|
| 325 |
"verified": true,
|
| 326 |
"ensemble": false,
|
| 327 |
"is_new": true,
|
| 328 |
"date_evaluated": "2026-07-16",
|
| 329 |
-
"model_url": "https://huggingface.co/
|
| 330 |
-
"overall":
|
| 331 |
-
"spatial":
|
| 332 |
-
"spatio_temporal":
|
| 333 |
-
"temporal":
|
| 334 |
-
"semantic":
|
| 335 |
-
"2d_localization":
|
| 336 |
-
"2d_referring_expressions":
|
| 337 |
-
"2d_spatial_pointing":
|
| 338 |
-
"single_object_tracking":
|
| 339 |
-
"temporal_localization":
|
| 340 |
-
"dense_video_captioning":
|
| 341 |
-
"event_verification":
|
| 342 |
-
"video_qa":
|
| 343 |
},
|
| 344 |
{
|
| 345 |
"id": "google-gemma-4-e2b-it",
|
|
@@ -353,97 +431,19 @@
|
|
| 353 |
"is_new": true,
|
| 354 |
"date_evaluated": "2026-07-16",
|
| 355 |
"model_url": "https://huggingface.co/google/gemma-4-e2b-it",
|
| 356 |
-
"overall": 21.
|
| 357 |
-
"spatial":
|
| 358 |
"spatio_temporal": 11.58,
|
| 359 |
-
"temporal": 9.
|
| 360 |
"semantic": 36.92,
|
| 361 |
-
"2d_localization":
|
| 362 |
"2d_referring_expressions": 5.52,
|
| 363 |
"2d_spatial_pointing": 31.74,
|
| 364 |
"single_object_tracking": 11.58,
|
| 365 |
"temporal_localization": 9.02,
|
| 366 |
-
"dense_video_captioning": 9.
|
| 367 |
"event_verification": 27.65,
|
| 368 |
"video_qa": 46.19
|
| 369 |
-
},
|
| 370 |
-
{
|
| 371 |
-
"id": "google-gemini-3-6-flash",
|
| 372 |
-
"name": "Gemini 3.6 Flash",
|
| 373 |
-
"organization": "Google",
|
| 374 |
-
"params": "—",
|
| 375 |
-
"type": "closed",
|
| 376 |
-
"result_type": "single",
|
| 377 |
-
"verified": true,
|
| 378 |
-
"ensemble": false,
|
| 379 |
-
"is_new": true,
|
| 380 |
-
"date_evaluated": "2026-07-21",
|
| 381 |
-
"model_url": "https://deepmind.google/technologies/gemini/",
|
| 382 |
-
"overall": 69.5187,
|
| 383 |
-
"spatial": 78.85,
|
| 384 |
-
"spatio_temporal": 75.99,
|
| 385 |
-
"temporal": 43.825,
|
| 386 |
-
"semantic": 79.41,
|
| 387 |
-
"2d_localization": 81.57,
|
| 388 |
-
"2d_referring_expressions": 75.78,
|
| 389 |
-
"2d_spatial_pointing": 79.2,
|
| 390 |
-
"single_object_tracking": 75.99,
|
| 391 |
-
"temporal_localization": 51.5,
|
| 392 |
-
"dense_video_captioning": 36.15,
|
| 393 |
-
"event_verification": 82,
|
| 394 |
-
"video_qa": 76.82
|
| 395 |
-
},
|
| 396 |
-
{
|
| 397 |
-
"id": "google-gemini-3-5-flash-lite",
|
| 398 |
-
"name": "Gemini 3.5 Flash-Lite",
|
| 399 |
-
"organization": "Google",
|
| 400 |
-
"params": "—",
|
| 401 |
-
"type": "closed",
|
| 402 |
-
"result_type": "single",
|
| 403 |
-
"verified": true,
|
| 404 |
-
"ensemble": false,
|
| 405 |
-
"is_new": true,
|
| 406 |
-
"date_evaluated": "2026-07-21",
|
| 407 |
-
"model_url": "https://deepmind.google/models/gemini/flash-lite/",
|
| 408 |
-
"overall": 52.8925,
|
| 409 |
-
"spatial": 61.09,
|
| 410 |
-
"spatio_temporal": 55.77,
|
| 411 |
-
"temporal": 26.7,
|
| 412 |
-
"semantic": 68.01,
|
| 413 |
-
"2d_localization": 65.73,
|
| 414 |
-
"2d_referring_expressions": 74.46,
|
| 415 |
-
"2d_spatial_pointing": 43.08,
|
| 416 |
-
"single_object_tracking": 55.77,
|
| 417 |
-
"temporal_localization": 24,
|
| 418 |
-
"dense_video_captioning": 29.4,
|
| 419 |
-
"event_verification": 70.41,
|
| 420 |
-
"video_qa": 65.61
|
| 421 |
-
},
|
| 422 |
-
{
|
| 423 |
-
"id": "openai-gpt-5-6-sol",
|
| 424 |
-
"name": "GPT-5.6 Sol",
|
| 425 |
-
"organization": "OpenAI",
|
| 426 |
-
"params": "—",
|
| 427 |
-
"type": "closed",
|
| 428 |
-
"result_type": "single",
|
| 429 |
-
"verified": true,
|
| 430 |
-
"ensemble": false,
|
| 431 |
-
"is_new": true,
|
| 432 |
-
"date_evaluated": "2026-07-09",
|
| 433 |
-
"model_url": "https://openai.com/index/gpt-5-6/",
|
| 434 |
-
"overall": 69.0258,
|
| 435 |
-
"spatial": 77.9333,
|
| 436 |
-
"spatio_temporal": 75.02,
|
| 437 |
-
"temporal": 46.25,
|
| 438 |
-
"semantic": 76.9,
|
| 439 |
-
"2d_localization": 69.3,
|
| 440 |
-
"2d_referring_expressions": 82.6,
|
| 441 |
-
"2d_spatial_pointing": 81.9,
|
| 442 |
-
"single_object_tracking": 75.02,
|
| 443 |
-
"temporal_localization": 54.8,
|
| 444 |
-
"dense_video_captioning": 37.7,
|
| 445 |
-
"event_verification": 75.9,
|
| 446 |
-
"video_qa": 77.9
|
| 447 |
}
|
| 448 |
]
|
| 449 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"schema_version": "1.0",
|
| 3 |
"benchmark_version": "1.0",
|
| 4 |
+
"updated": "2026-09-10",
|
| 5 |
"models": [
|
| 6 |
+
{
|
| 7 |
+
"id": "google-gemini-3-6-flash",
|
| 8 |
+
"name": "Gemini 3.6 Flash",
|
| 9 |
+
"organization": "Google",
|
| 10 |
+
"params": "—",
|
| 11 |
+
"type": "closed",
|
| 12 |
+
"result_type": "single",
|
| 13 |
+
"verified": true,
|
| 14 |
+
"ensemble": false,
|
| 15 |
+
"is_new": true,
|
| 16 |
+
"date_evaluated": "2026-07-21",
|
| 17 |
+
"model_url": "https://deepmind.google/technologies/gemini/",
|
| 18 |
+
"overall": 69.52,
|
| 19 |
+
"spatial": 78.85,
|
| 20 |
+
"spatio_temporal": 75.99,
|
| 21 |
+
"temporal": 43.83,
|
| 22 |
+
"semantic": 79.41,
|
| 23 |
+
"2d_localization": 81.57,
|
| 24 |
+
"2d_referring_expressions": 75.78,
|
| 25 |
+
"2d_spatial_pointing": 79.2,
|
| 26 |
+
"single_object_tracking": 75.99,
|
| 27 |
+
"temporal_localization": 51.5,
|
| 28 |
+
"dense_video_captioning": 36.15,
|
| 29 |
+
"event_verification": 82.0,
|
| 30 |
+
"video_qa": 76.82
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"id": "openai-gpt-5-6-sol",
|
| 34 |
+
"name": "GPT-5.6 Sol",
|
| 35 |
+
"organization": "OpenAI",
|
| 36 |
+
"params": "—",
|
| 37 |
+
"type": "closed",
|
| 38 |
+
"result_type": "single",
|
| 39 |
+
"verified": true,
|
| 40 |
+
"ensemble": false,
|
| 41 |
+
"is_new": true,
|
| 42 |
+
"date_evaluated": "2026-07-09",
|
| 43 |
+
"model_url": "https://openai.com/index/gpt-5-6/",
|
| 44 |
+
"overall": 68.04,
|
| 45 |
+
"spatial": 73.55,
|
| 46 |
+
"spatio_temporal": 75.02,
|
| 47 |
+
"temporal": 46.5,
|
| 48 |
+
"semantic": 77.11,
|
| 49 |
+
"2d_localization": 69.3,
|
| 50 |
+
"2d_referring_expressions": 69.44,
|
| 51 |
+
"2d_spatial_pointing": 81.9,
|
| 52 |
+
"single_object_tracking": 75.02,
|
| 53 |
+
"temporal_localization": 55.71,
|
| 54 |
+
"dense_video_captioning": 37.28,
|
| 55 |
+
"event_verification": 76.14,
|
| 56 |
+
"video_qa": 78.08
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"id": "nvidia-cosmos3-super",
|
| 60 |
+
"name": "Cosmos3-Super",
|
| 61 |
+
"organization": "NVIDIA",
|
| 62 |
+
"params": "64B",
|
| 63 |
+
"type": "open",
|
| 64 |
+
"result_type": "single",
|
| 65 |
+
"verified": true,
|
| 66 |
+
"ensemble": false,
|
| 67 |
+
"is_new": true,
|
| 68 |
+
"date_evaluated": "2026-05-29",
|
| 69 |
+
"model_url": "https://huggingface.co/nvidia/Cosmos3-Super",
|
| 70 |
+
"overall": 63.62,
|
| 71 |
+
"spatial": 78.71,
|
| 72 |
+
"spatio_temporal": 64.66,
|
| 73 |
+
"temporal": 40.72,
|
| 74 |
+
"semantic": 70.37,
|
| 75 |
+
"2d_localization": 86.97,
|
| 76 |
+
"2d_referring_expressions": 76.23,
|
| 77 |
+
"2d_spatial_pointing": 72.94,
|
| 78 |
+
"single_object_tracking": 64.66,
|
| 79 |
+
"temporal_localization": 51.9,
|
| 80 |
+
"dense_video_captioning": 29.54,
|
| 81 |
+
"event_verification": 71.28,
|
| 82 |
+
"video_qa": 69.46
|
| 83 |
+
},
|
| 84 |
{
|
| 85 |
"id": "google-gemini-3-1-pro",
|
| 86 |
"name": "Gemini 3.1 Pro",
|
|
|
|
| 93 |
"is_new": false,
|
| 94 |
"date_evaluated": "2026-05-27",
|
| 95 |
"model_url": "https://deepmind.google/technologies/gemini/",
|
| 96 |
+
"overall": 62.66,
|
| 97 |
+
"spatial": 72.14,
|
| 98 |
+
"spatio_temporal": 67.88,
|
| 99 |
+
"temporal": 40.62,
|
| 100 |
+
"semantic": 70.01,
|
| 101 |
+
"2d_localization": 77.21,
|
| 102 |
"2d_referring_expressions": 67.17,
|
| 103 |
+
"2d_spatial_pointing": 72.04,
|
| 104 |
+
"single_object_tracking": 67.88,
|
| 105 |
"temporal_localization": 45.69,
|
| 106 |
+
"dense_video_captioning": 35.55,
|
| 107 |
+
"event_verification": 68.57,
|
| 108 |
+
"video_qa": 71.46
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"id": "nvidia-cosmos3-nano",
|
| 112 |
+
"name": "Cosmos3-Nano",
|
| 113 |
+
"organization": "NVIDIA",
|
| 114 |
+
"params": "16B",
|
| 115 |
+
"type": "open",
|
| 116 |
+
"result_type": "single",
|
| 117 |
+
"verified": true,
|
| 118 |
+
"ensemble": false,
|
| 119 |
+
"is_new": true,
|
| 120 |
+
"date_evaluated": "2026-05-29",
|
| 121 |
+
"model_url": "https://huggingface.co/nvidia/Cosmos3-Nano",
|
| 122 |
+
"overall": 60.67,
|
| 123 |
+
"spatial": 74.84,
|
| 124 |
+
"spatio_temporal": 59.21,
|
| 125 |
+
"temporal": 39.73,
|
| 126 |
+
"semantic": 68.91,
|
| 127 |
+
"2d_localization": 74.11,
|
| 128 |
+
"2d_referring_expressions": 75.57,
|
| 129 |
+
"2d_spatial_pointing": 74.83,
|
| 130 |
+
"single_object_tracking": 59.21,
|
| 131 |
+
"temporal_localization": 48.04,
|
| 132 |
+
"dense_video_captioning": 31.42,
|
| 133 |
+
"event_verification": 68.88,
|
| 134 |
+
"video_qa": 68.95
|
| 135 |
},
|
| 136 |
{
|
| 137 |
"id": "alibaba-qwen3-vl-32b",
|
|
|
|
| 145 |
"is_new": false,
|
| 146 |
"date_evaluated": "2026-05-27",
|
| 147 |
"model_url": "https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct",
|
| 148 |
+
"overall": 55.38,
|
| 149 |
+
"spatial": 73.56,
|
| 150 |
"spatio_temporal": 44.19,
|
| 151 |
+
"temporal": 38.1,
|
| 152 |
"semantic": 65.67,
|
| 153 |
+
"2d_localization": 72.67,
|
| 154 |
"2d_referring_expressions": 72.39,
|
| 155 |
"2d_spatial_pointing": 75.62,
|
| 156 |
"single_object_tracking": 44.19,
|
| 157 |
+
"temporal_localization": 46.8,
|
| 158 |
"dense_video_captioning": 29.4,
|
| 159 |
"event_verification": 60.04,
|
| 160 |
"video_qa": 71.3
|
|
|
|
| 171 |
"is_new": false,
|
| 172 |
"date_evaluated": "2026-05-27",
|
| 173 |
"model_url": "https://huggingface.co/nvidia/Cosmos-Reason2-8B",
|
| 174 |
+
"overall": 54.46,
|
| 175 |
+
"spatial": 74.25,
|
| 176 |
"spatio_temporal": 37.69,
|
| 177 |
"temporal": 39.9,
|
| 178 |
"semantic": 66.02,
|
| 179 |
"2d_localization": 83.88,
|
| 180 |
+
"2d_referring_expressions": 70.26,
|
| 181 |
"2d_spatial_pointing": 68.6,
|
| 182 |
"single_object_tracking": 37.69,
|
| 183 |
"temporal_localization": 47.3,
|
|
|
|
| 185 |
"event_verification": 64.09,
|
| 186 |
"video_qa": 67.95
|
| 187 |
},
|
| 188 |
+
{
|
| 189 |
+
"id": "google-gemini-3-5-flash-lite",
|
| 190 |
+
"name": "Gemini 3.5 Flash-Lite",
|
| 191 |
+
"organization": "Google",
|
| 192 |
+
"params": "—",
|
| 193 |
+
"type": "closed",
|
| 194 |
+
"result_type": "single",
|
| 195 |
+
"verified": true,
|
| 196 |
+
"ensemble": false,
|
| 197 |
+
"is_new": true,
|
| 198 |
+
"date_evaluated": "2026-07-21",
|
| 199 |
+
"model_url": "https://deepmind.google/models/gemini/flash-lite/",
|
| 200 |
+
"overall": 53.25,
|
| 201 |
+
"spatial": 61.09,
|
| 202 |
+
"spatio_temporal": 55.77,
|
| 203 |
+
"temporal": 28.14,
|
| 204 |
+
"semantic": 68.01,
|
| 205 |
+
"2d_localization": 65.73,
|
| 206 |
+
"2d_referring_expressions": 74.46,
|
| 207 |
+
"2d_spatial_pointing": 43.08,
|
| 208 |
+
"single_object_tracking": 55.77,
|
| 209 |
+
"temporal_localization": 24.0,
|
| 210 |
+
"dense_video_captioning": 32.29,
|
| 211 |
+
"event_verification": 70.41,
|
| 212 |
+
"video_qa": 65.61
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"id": "google-gemini-3-2-flash-lite",
|
| 216 |
+
"name": "Gemini 3.1 Flash Lite",
|
| 217 |
+
"organization": "Google",
|
| 218 |
+
"params": "—",
|
| 219 |
+
"type": "closed",
|
| 220 |
+
"result_type": "single",
|
| 221 |
+
"verified": true,
|
| 222 |
+
"ensemble": false,
|
| 223 |
+
"is_new": false,
|
| 224 |
+
"date_evaluated": "2026-05-27",
|
| 225 |
+
"model_url": "https://deepmind.google/technologies/gemini/",
|
| 226 |
+
"overall": 52.61,
|
| 227 |
+
"spatial": 62.71,
|
| 228 |
+
"spatio_temporal": 46.68,
|
| 229 |
+
"temporal": 35.13,
|
| 230 |
+
"semantic": 65.91,
|
| 231 |
+
"2d_localization": 71.74,
|
| 232 |
+
"2d_referring_expressions": 52.7,
|
| 233 |
+
"2d_spatial_pointing": 63.68,
|
| 234 |
+
"single_object_tracking": 46.68,
|
| 235 |
+
"temporal_localization": 37.72,
|
| 236 |
+
"dense_video_captioning": 32.54,
|
| 237 |
+
"event_verification": 63.78,
|
| 238 |
+
"video_qa": 68.03
|
| 239 |
+
},
|
| 240 |
{
|
| 241 |
"id": "alibaba-qwen3-vl-8b",
|
| 242 |
"name": "Qwen3-VL-8B-Instruct",
|
|
|
|
| 249 |
"is_new": false,
|
| 250 |
"date_evaluated": "2026-05-27",
|
| 251 |
"model_url": "https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct",
|
| 252 |
+
"overall": 50.07,
|
| 253 |
"spatial": 67.26,
|
| 254 |
"spatio_temporal": 33.12,
|
| 255 |
"temporal": 36.98,
|
| 256 |
+
"semantic": 62.91,
|
| 257 |
"2d_localization": 59.93,
|
| 258 |
"2d_referring_expressions": 73.29,
|
| 259 |
"2d_spatial_pointing": 68.56,
|
|
|
|
| 275 |
"is_new": false,
|
| 276 |
"date_evaluated": "2026-05-27",
|
| 277 |
"model_url": "https://huggingface.co/Qwen/Qwen3.5-VL-27B-Instruct",
|
| 278 |
+
"overall": 49.25,
|
| 279 |
"spatial": 79.04,
|
| 280 |
"spatio_temporal": 24.4,
|
| 281 |
"temporal": 31.93,
|
|
|
|
| 301 |
"is_new": false,
|
| 302 |
"date_evaluated": "2026-05-27",
|
| 303 |
"model_url": "https://huggingface.co/Qwen/Qwen3.5-VL-9B-Instruct",
|
| 304 |
+
"overall": 48.3,
|
| 305 |
"spatial": 75.26,
|
| 306 |
"spatio_temporal": 23.52,
|
| 307 |
+
"temporal": 35.2,
|
| 308 |
+
"semantic": 59.2,
|
| 309 |
"2d_localization": 76.87,
|
| 310 |
"2d_referring_expressions": 77.16,
|
| 311 |
"2d_spatial_pointing": 71.74,
|
| 312 |
"single_object_tracking": 23.52,
|
| 313 |
"temporal_localization": 42.53,
|
| 314 |
+
"dense_video_captioning": 27.88,
|
| 315 |
"event_verification": 46.78,
|
| 316 |
"video_qa": 71.63
|
| 317 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 318 |
{
|
| 319 |
"id": "nvidia-cosmos-reason2-32b",
|
| 320 |
"name": "Cosmos-Reason2-32B",
|
|
|
|
| 327 |
"is_new": false,
|
| 328 |
"date_evaluated": "2026-05-27",
|
| 329 |
"model_url": "https://huggingface.co/nvidia/Cosmos-Reason2-32B",
|
| 330 |
+
"overall": 47.03,
|
| 331 |
+
"spatial": 42.81,
|
| 332 |
"spatio_temporal": 33.12,
|
| 333 |
+
"temporal": 40.25,
|
| 334 |
"semantic": 71.94,
|
| 335 |
+
"2d_localization": 4.35,
|
| 336 |
"2d_referring_expressions": 50.04,
|
| 337 |
"2d_spatial_pointing": 74.03,
|
| 338 |
"single_object_tracking": 33.12,
|
|
|
|
| 342 |
"video_qa": 70.29
|
| 343 |
},
|
| 344 |
{
|
| 345 |
+
"id": "nvidia-cosmos3-edge",
|
| 346 |
+
"name": "Cosmos3-Edge",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 347 |
"organization": "NVIDIA",
|
| 348 |
+
"params": "4B",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 349 |
"type": "open",
|
| 350 |
"result_type": "single",
|
| 351 |
"verified": true,
|
| 352 |
"ensemble": false,
|
| 353 |
"is_new": true,
|
| 354 |
"date_evaluated": "2026-07-16",
|
| 355 |
+
"model_url": "https://huggingface.co/nvidia/Cosmos3-Edge",
|
| 356 |
+
"overall": 46.94,
|
| 357 |
+
"spatial": 72.02,
|
| 358 |
+
"spatio_temporal": 24.65,
|
| 359 |
+
"temporal": 28.07,
|
| 360 |
+
"semantic": 63.02,
|
| 361 |
+
"2d_localization": 79.02,
|
| 362 |
+
"2d_referring_expressions": 69.27,
|
| 363 |
+
"2d_spatial_pointing": 67.76,
|
| 364 |
+
"single_object_tracking": 24.65,
|
| 365 |
+
"temporal_localization": 39.51,
|
| 366 |
+
"dense_video_captioning": 16.64,
|
| 367 |
+
"event_verification": 63.36,
|
| 368 |
+
"video_qa": 62.68
|
| 369 |
},
|
| 370 |
{
|
| 371 |
"id": "nvidia-cosmos-reason2-2b",
|
|
|
|
| 379 |
"is_new": true,
|
| 380 |
"date_evaluated": "2026-07-16",
|
| 381 |
"model_url": "https://huggingface.co/nvidia/Cosmos-Reason2-2B",
|
| 382 |
+
"overall": 45.94,
|
| 383 |
+
"spatial": 63.2,
|
| 384 |
"spatio_temporal": 26.85,
|
| 385 |
+
"temporal": 33.73,
|
| 386 |
"semantic": 59.98,
|
| 387 |
+
"2d_localization": 73.56,
|
| 388 |
"2d_referring_expressions": 56.33,
|
| 389 |
"2d_spatial_pointing": 59.7,
|
| 390 |
"single_object_tracking": 26.85,
|
|
|
|
| 394 |
"video_qa": 64.69
|
| 395 |
},
|
| 396 |
{
|
| 397 |
+
"id": "alibaba-qwen3-vl-2b",
|
| 398 |
+
"name": "Qwen3-VL-2B-Instruct",
|
| 399 |
+
"organization": "Alibaba",
|
| 400 |
+
"params": "2B",
|
| 401 |
"type": "open",
|
| 402 |
"result_type": "single",
|
| 403 |
"verified": true,
|
| 404 |
"ensemble": false,
|
| 405 |
"is_new": true,
|
| 406 |
"date_evaluated": "2026-07-16",
|
| 407 |
+
"model_url": "https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct",
|
| 408 |
+
"overall": 41.86,
|
| 409 |
+
"spatial": 55.22,
|
| 410 |
+
"spatio_temporal": 30.38,
|
| 411 |
+
"temporal": 27.55,
|
| 412 |
+
"semantic": 54.29,
|
| 413 |
+
"2d_localization": 47.42,
|
| 414 |
+
"2d_referring_expressions": 65.12,
|
| 415 |
+
"2d_spatial_pointing": 53.13,
|
| 416 |
+
"single_object_tracking": 30.38,
|
| 417 |
+
"temporal_localization": 35.21,
|
| 418 |
+
"dense_video_captioning": 19.9,
|
| 419 |
+
"event_verification": 44.73,
|
| 420 |
+
"video_qa": 63.85
|
| 421 |
},
|
| 422 |
{
|
| 423 |
"id": "google-gemma-4-e2b-it",
|
|
|
|
| 431 |
"is_new": true,
|
| 432 |
"date_evaluated": "2026-07-16",
|
| 433 |
"model_url": "https://huggingface.co/google/gemma-4-e2b-it",
|
| 434 |
+
"overall": 21.28,
|
| 435 |
+
"spatial": 27.55,
|
| 436 |
"spatio_temporal": 11.58,
|
| 437 |
+
"temporal": 9.08,
|
| 438 |
"semantic": 36.92,
|
| 439 |
+
"2d_localization": 45.4,
|
| 440 |
"2d_referring_expressions": 5.52,
|
| 441 |
"2d_spatial_pointing": 31.74,
|
| 442 |
"single_object_tracking": 11.58,
|
| 443 |
"temporal_localization": 9.02,
|
| 444 |
+
"dense_video_captioning": 9.14,
|
| 445 |
"event_verification": 27.65,
|
| 446 |
"video_qa": 46.19
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 447 |
}
|
| 448 |
]
|
| 449 |
}
|