Which model leads OCRBenchv2?

Q: Which model leads OCRBenchv2?

Qwen3.6 27B tops the board at 68.3

OCRBenchv2 Leaderboard — Model Scores

Pro Mode — raw scores, one benchmark under the lensthird-party eval harness

The field · 36 measured

38.168.3

36 models measured, most on the third-party eval harness. Qwen3.6 27B tops the board at 68.3.

36 measured·3 new this week·Lifecycle Updated Oct 8, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only24 of 36Bars show 38.1–68.3 · ◆ solid = first-party or better · outlined = arm's-length

01Qwen3.6 27BNEWAlibaba68.3$0.60/M$3.6/M$0.031

02Qwen3.7 Plus PreviewAlibaba67.1VENDOR$0.40/M$1.6/M$0.015

03Nemotron 3 Nano Omni+1 altNVIDIA65.8$0.30/M$0.90/M$0.009

04Qwen3.6 35B A3BAlibaba65.5$0.38/M$2.3/M$0.020

05Qwen3.5 35B A3BAlibaba65.3$0.25/M$2.0/M$0.017

06Qwen3.8 27BNEWAlibaba63.5$0.50/M$3.0/M$0.028

07Gemini 3 ProGoogle63.4VENDOR$2.0/M$12.0/M$0.110

08Seed 2.1 Pro Preview+1 altByteDance63.2VENDOR———

09Seed2.1ByteDance62.8VENDOR———

10Gemini 3.1 ProGoogle62.8VENDOR$2.0/M$12.0/M$0.111

11Seed1.6 VisionByteDance62.2———

12Qwen3 Omni 30B A3B InstructAlibaba61.3VENDOR$0.25/M$0.97/M$0.010

13NVIDIA Nemotron Nano 12B v2 VL+1 altNVIDIA61.2VENDOR$0.20/M$0.60/M$0.007

14GPT-5.5OpenAI61.1VENDOR$5.0/M$30.0/M$0.286

15Gemini 2.5 ProGoogle59.3$1.3/M$10.0/M$0.095

16Qwen2.5 VL 32B InstructAlibaba59.1VENDOR———

17GLM-4.6V-Flash (9B)Z.ai59$0.30/M$0.90/M$0.010

18Qwen3.5 9BAlibaba58.7$0.14/M$0.20/M$0.003

19Qwen2.5 Omni 7BAlibaba57.8VENDOR$0.10/M$6.8/M$0.059

20Muse GlimmerNEWMeta56.9$0.33/M$1.4/M$0.015

21Claude Opus 4.7Anthropic56.9VENDOR$5.0/M$25.0/M$0.264

22Llama 3.1 Nemotron Nano VL 8B V1NVIDIA56.4———

23GPT-5+2 altsOpenAI55.5$1.3/M$10.0/M$0.101

24Gemma 4 12BGoogle51.8$0.10/M$0.30/M$0.004

25Gemini 1.5 ProGoogle51.6———

26GPT-5.2OpenAI50.5$1.8/M$14.0/M$0.156

27Claude Opus 4.6Anthropic48.4$5.0/M$25.0/M$0.310

28LFM2.5-VL-3B+1 altLiquid AI47.5VENDOR———

29Qwen2 VL 72B InstructAlibaba46.1VENDOR———

30MiniCPM-o-4.5OpenBMB44.9VENDOR———

31Molmo2-8BAllenAI44.7VENDOR———

32Gemma 4 E2BGoogle44.7———

33LFM2-VL-3BLiquid AI43.9———

34Ministral 3 14BMistral43.3$0.20/M$0.20/M$0.005

35MiniCPM-V 4.6OpenBMB40.4———

36Phi 4 Multimodal Instruct+1 altMicrosoft38.1VENDOR———

◆ 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: 36 models scored · 24 independently verified · 5 vendor cross-reference · 7 vendor-reported · 0 with source disagreement. How these tiers are assigned