Which model leads MV-Bench?

Q: Which model leads MV-Bench?

GLM 5.3 Flash tops the board at 77.8

MV-Bench Leaderboard — 22 Models Ranked

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

The field · 20 measured

44.977.8

20 models measured, most on the official card harness. GLM 5.3 Flash tops the board at 77.8.

20 measured·Lifecycle Updated Oct 8, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only0 of 20Bars show 44.9–77.8 · ◆ solid = first-party or better · outlined = arm's-length

01GLM 5.3 FlashZ.ai77.8$0.15/M$0.50/M$0.004

02Qwen3.5 122B A10BAlibaba76.6$0.40/M$3.2/M$0.023

03Qwen3.6 27BAlibaba75.5$0.60/M$3.6/M$0.028

04Qwen3.5 35B A3BAlibaba74.8$0.25/M$2.0/M$0.015

05Qwen3.5 27BAlibaba74.6$0.30/M$2.4/M$0.018

06Qwen3.6 35B A3BAlibaba74.6$0.38/M$2.3/M$0.018

07Qwen2 VL 72B Instruct+1 altAlibaba73.6———

08Qwen3 VL 32B ReasoningAlibaba73.2$0.16/M$0.64/M$0.005

09Qwen3 VL 32B InstructAlibaba72.8$0.16/M$0.64/M$0.005

10Qwen3 VL 30B A3B InstructAlibaba72.3$0.20/M$0.80/M$0.007

11Qwen3 VL 30B A3B ReasoningAlibaba72$0.20/M$2.4/M$0.018

12Qwen2.5 VL 72BAlibaba70.4———

13Qwen2.5 Omni 7BAlibaba70.3$0.10/M$6.8/M$0.049

14Qwen3 VL 4B (Reasoning)+1 altAlibaba69.3———

15Qwen3 VL Thinking (8B)Alibaba69$0.18/M$2.1/M$0.017

16Qwen3 VL 4B InstructAlibaba68.9———

17Qwen3 VL 8B InstructAlibaba68.7$0.18/M$0.70/M$0.006

18Mage-VL-4BMicrosoft65.13P———

19Phi 4 R V 15BMicrosoft49.23P———

20Phi 4 Multimodal InstructMicrosoft44.93P———

◆ 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.

Verification: 20 models scored · 0 independently verified · 20 vendor-reported · 0 with source disagreement. How these tiers are assigned