List your company

Multi-image reasoning on Mantis (test)

72.81AccuracyDelimScaling
29.639640.847352.05563.2627Sep 30, 2024Dec 20, 2024Mar 12, 2025Jun 2, 2025Aug 22, 2025Nov 12, 2025Feb 2, 2026

Evaluation results

MethodLinks
DelimScalingBackbone=InternVL3, Model Size=14B2026.0272.81
InternVL3Model Size=14B, Method Variant=Baseline2026.0271.89
DelimScalingBackbone=Qwen2.5-VL, Model Size=32B2026.0270.05
DelimScalingBackbone=Qwen2.5-VL, Model Size=7B2026.0269.12
DelimScalingBackbone=InternVL3, Model Size=8B2026.0269.12
Qwen2.5-VLModel Size=7B, Method Variant=Baseline2026.0268.66
Qwen2.5-VLModel Size=32B, Method Variant=Baseline2026.0268.2
InternVL3Model Size=8B, Method Variant=Baseline2026.0267.28
LLaVA-NeXT-Interleave-14BModel Scale=>14B2024.0966.4
MM1.5-30BModel Scale=>14B2024.0964.6
DelimScalingBackbone=LLaVA-OV, Model Size=7B2026.0264.06
DelimScalingBackbone=Qwen2.5-VL, Model Size=3B2026.0263.13
LLaVA-NeXT-Interleave-7BModel Scale=7B2024.0962.7
GPT-4VModel Scale=>14B2024.0962.7
LLaVA-OVModel Size=7B, Method Variant=Baseline2026.0262.21
Qwen2.5-VLModel Size=3B, Method Variant=Baseline2026.0259.91
MM1.5-7BModel Scale=7B2024.0957.6
Mantis-Idefics2-8BModel Scale=7B2024.0957.1
BLIP-3-4BModel Scale=3B2024.0956.7
MM1.5-3BModel Scale=3B2024.0954.8
MM1.5-3B-MoEModel Scale=3B2024.0954.4
DelimScalingBackbone=InternVL3, Model Size=2B2026.0254.38
MM1-30BModel Scale=>14B2024.0952.5
InternVL3Model Size=2B, Method Variant=Baseline2026.0252.07
MM1-7BModel Scale=7B2024.0951.6
MM1.5-1B-MoEModel Scale=1B2024.0951.2
MM1.5-1BModel Scale=1B2024.0950.7
DelimScalingBackbone=InternVL3, Model Size=1B2026.0249.77
Idefics2-8BModel Scale=7B2024.0948.9
Phi-3-Vision-4BModel Scale=3B2024.0947.9
InternVL3Model Size=1B, Method Variant=Baseline2026.0247
LLaVA-NeXT-Interleave-0.5BModel Scale=1B2024.0945.6
MM1-3BModel Scale=3B2024.0945.2
MM1-1BModel Scale=1B2024.0941.5
DelimScalingBackbone=LLaVA-OV, Model Size=0.5B2026.0241.01
LLaVA-OVModel Size=0.5B, Method Variant=Baseline2026.0240.09
LLaVAOneVision-0.5BModel Scale=1B2024.0939.6
Emu2-Chat-37BModel Scale=>14B2024.0937.8
LLaVA-v1.5-7BModel Scale=7B2024.0931.3