context.vn

Cross-Domain Efficiency Charts

3,378 models across 7 domains · Back to table · API

Models by Domain

Vision1,615 (48%)3D771 (23%)Embedding428 (13%)Multimodal299 (9%)Code125 (4%)Speech91 (3%)Video49 (1%)

Parameter Size Distribution by Domain

Number of models in each parameter bucket, grouped by domain (top 6 domains shown)

02815628431124<1M1-10M10-100M100M-1B>1BVision3DEmbeddingMultimodalCodeSpeech

Parameters vs Score by Domain

Each dot is a model. X-axis: parameter count (log scale). Y-axis: primary benchmark score.

Code: Params vs HumanEval+Parameters (log scale)HumanEval+1.823.244.565.987.21.0B
Multimodal: Params vs MMBenchParameters (log scale)MMBench1.123.045.066.988.81.0B
Video: Params vs MVBenchParameters (log scale)MVBench33.044.556.167.779.21.0B
Vision: Params vs mAPParameters (log scale)mAP0.422.845.267.690.110M100M

Top Performers by Task

CodeCode Generation

1Qwen2.5-Coder-32B-Instruct32.0B87.2
2DeepSeek-V3 (Nov 2024)37.0B86.6
3DeepSeek-V2.5 (Nov 2024)21.0B83.5
4DeepSeek-Coder-V2-Instruct21.0B82.3
5CodeQwen1.5-7B-Chat7.0B78.7
Metric: HumanEval+

MultimodalVision-Language

1InternVL3-78B78.4B88.8
2InternVL2.5-78B-MPO78.0B88.4
3Qwen2.5-VL-72B73.4B88.3
4InternVL2.5-78B78.0B88.2
5InternVL3-38B38.4B87.7
Metric: MMBench

VideoVideo Understanding

1InternVL3-78B78.4B79.2
2InternVL3-38B38.4B76.0
3InternVL2_5-78B78.4B75.6
4InternVL3-8B7.9B73.2
5Qwen2.5-VL-72B73.4B71.3
Metric: MVBench

VisionObject Detection

1D-FINE-X-Objects365+COCO62M0.6
2RF-DETR-L136M0.6
3LW-DETR-xlarge118M0.6
4D-FINE-L-Objects365+COCO31M0.6
5DEIM-D-FINE-X62M0.6
Metric: mAP

VisionImage Classification

1eva02_large_patch14_448.mim_m38m_ft_in22k_in1k305M90.1
2eva02_large_patch14_448.mim_in22k_ft_in22k_in1k305M90.0
3eva02_large_patch14_448.mim_in22k_ft_in1k305M89.6
4eva02_large_patch14_448.mim_m38m_ft_in1k305M89.5
5vit_so400m_patch14_siglip_378.webli_ft_in1k429M89.4
Metric: Top-1