Which model leads MMBench?

Step3-VL-10B leads the MMBench board of 9 measured models on the Vector Wire tracker at 91.8. Every score is source-attributed with a verification tier.

MMBench Leaderboard — Model Scores

Refresh Models —Deals —Regulatory —Sources —

Pro Mode — raw scores, one benchmark under the lensofficial card harness

The field · 9 measured

8091.8

9 models measured, most on the official card harness. Step3-VL-10B tops the board at 91.8.

9 measured·9 new this week·Lifecycle Updated Sep 6, 2026Basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only0 of 9Bars show 80–91.8· ◆ solid = first-party or better · outlined = arm's-length

01Step3-VL-10BNEWStepFun91.8———

02Qwen2.5 VL 72B InstructNEWAlibaba88———

03Qwen3.5-4BNEWAlibaba87.13P$0.03/M$0.15/M$0.001

04Phi-4-multimodal-instructNEWMicrosoft86.7———

05Qwen2-VL-72B-InstructNEWAlibaba86.5———

06Qwen-Drive-1.0-SFTNEWAlibaba85.53P———

07Phi-3.5-vision-instructNEWMicrosoft81.9———

08LFM2.5-VL-3BNEWLiquid AI81———

09InternVL3.5-8B-InstructNEWOpenGVLab803P———

◆ solid = independently verified · aggregator attested · vendor attributed — outlined = cross-reference · source-attributed. Dimmed ◆ = confidence inferred. “+N alts” = alternate disclosures in the drawer with why-demoted. Prices per million tokens; $/point = input price ÷ score. Every row opens its model dossier.