OCRBench v2
The official leaderboard hasn’t scored a model released after Nemotron 3.5 Super VL (October 2026). Labs report their own numbers for one newer model: MiniCPM-OCR-1.0 at 54.7%.
Visual-text localization and reasoning tasks across 31 scenarios (documents, scenes, charts). Averaged task accuracy.
- Made by
- HUST (Yuliang Liu et al.)
- Released
- December 2024
- Kind
- Documents, charts and text in images
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- Public, plus a private test set
- Measures
- Capability
- Interaction
- Single turn
- Refresh
- New versions replace it
- Run by
- A host
- Based on
- OCRBench
- In views
- Vision
- Strongest model tested
-
Qwen 3.8 Max #30 on the capability index
- Last in a launch post
- October 2026
- Last model added
- October 2026
- Scores refreshed
- November 18th, 2026
In the authors’ words
Scoring the Optical Character Recognition (OCR) capabilities of Large Multimodal Models (LMMs) has witnessed growing interest. Existing benchmarks have highlighted the impressive performance of LMMs in text recognition; however, their abilities in certain challenging tasks, such as text localization, handwritten content extraction, and logical reasoning, remain underexplored. To bridge this gap, we introduce OCRBench v2, a large-scale bilingual text-centric benchmark with currently the most comprehensive set of tasks (4x more tasks than the previous multi-scene benchmark OCRBench), the widest coverage of scenarios (31 diverse scenarios), and thorough evaluation metrics, with 10,000 human-verified question-answering pairs and a high proportion of difficult samples. Moreover, we construct a private test set with 1,500 manually annotated images. The consistent evaluation trends observed across both public and private test sets validate the OCRBench v2's reliability. After carefully benchmarking state-of-the-art LMMs, we find that most LMMs score below 50 (100 in total) and suffer from five-type limitations, including less frequently encountered text recognition, fine-grained perception, layout perception, complex element parsing, and logical reasoning. The project website is at: https://99franklin.github.io/ocrbench_v2/
KDL Frontier (2026-10, EN) tops the leaderboard with 76.9% Qwen3.6 27B leads with 68.3%
Best score per entry from the benchmark's own leaderboard, read on November 18th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Qwen3.6 27B, at 68.3%. This view shows only models we track, each at its best.
Qwen 3.8 Max leads with 74.2%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.
- 76.9 KDL Frontier (2026-10, EN)
- 76.2 SEEU DOC
- 73.4 Yaochi-DTS-A2T-1.9
- 71.7 U2-OCR
- 69.8 Posicube_DVLM
- 68.3
Qwen3.6 27B
- 68.1 KDL Frontier
- 67.4
Nemotron 3.5 Super VL
- 65.8
Nemotron 3 Nano Omni 30B-A3B
- 65.5
Qwen 3.6 35B-A3B
- 65.3
Qwen3.5-35B-A3B
- 63.5
Qwen 3.8 27B
- 63.4
Gemini 3 Pro
- 62.2 Seed1.6-vision
- 61.8 TeleMM-2.0
- 61.3
Qwen3-Omni-30B-A3B
- 61.2
Nemotron Nano 12B v2 VL
- 59.9 SenseNova-U1.5-8B-MoT
- 59.3 Gemini-2.5-Pro (snapshot unspecified)
- 59.0 GLM-4.6V-Flash
- 58.7
Qwen3.5-9B
- 58.2 SenseNova-U1
- 56.9
Muse Glimmer
- 56.4
Llama-3.1-Nemotron-Nano-VL-8B-V1
- 55.5 GPT-5
- 42.6 LFM2.5-VL-3B
- 1 KDL Frontier (2026-10, EN) 76.9%
- 2 SEEU DOC 76.2%
- 3 Yaochi-DTS-A2T-1.9 73.4%
- 4 U2-OCR 71.7%
- 5 Posicube_DVLM 69.8%
- 6
Qwen3.6 27B 68.3%
- 7 KDL Frontier 68.1%
- 8
Nemotron 3.5 Super VL 67.4%
- 9
Nemotron 3 Nano Omni 30B-A3B 65.8%
- 10
Qwen 3.6 35B-A3B 65.5%
- 11
Qwen3.5-35B-A3B 65.3%
- 12
Qwen 3.8 27B 63.5%
- 13
Gemini 3 Pro 63.4%
- 14 Seed1.6-vision 62.2%
- 15 TeleMM-2.0 61.8%
- 16
Qwen3-Omni-30B-A3B 61.3%
- 17
Nemotron Nano 12B v2 VL 61.2%
- 18 SenseNova-U1.5-8B-MoT 59.9%
- 19 Gemini-2.5-Pro (snapshot unspecified) 59.3%
- 20 GLM-4.6V-Flash 59%
- 21
Qwen3.5-9B 58.7%
- 22 SenseNova-U1 58.2%
- 23
Muse Glimmer 56.9%
- 24
Llama-3.1-Nemotron-Nano-VL-8B-V1 56.4%
- 25 GPT-5 55.5%
- 26 LFM2.5-VL-3B 42.6%
Source: Official OCRBench v2 leaderboard
- 68.3
Qwen3.6 27B
- 67.4
Nemotron 3.5 Super VL
- 65.8
Nemotron 3 Nano Omni 30B-A3B
- 65.5
Qwen 3.6 35B-A3B
- 65.3
Qwen3.5-35B-A3B
- 63.5
Qwen 3.8 27B
- 63.4
Gemini 3 Pro
- 61.3
Qwen3-Omni-30B-A3B
- 61.2
Nemotron Nano 12B v2 VL
- 59.0 GLM-4.6V-Flash
- 58.7
Qwen3.5-9B
- 56.9
Muse Glimmer
- 56.4
Llama-3.1-Nemotron-Nano-VL-8B-V1
- 55.5 GPT-5
- 42.6 LFM2.5-VL-3B
- 1
Qwen3.6 27B 68.3%
- 2
Nemotron 3.5 Super VL 67.4%
- 3
Nemotron 3 Nano Omni 30B-A3B 65.8%
- 4
Qwen 3.6 35B-A3B 65.5%
- 5
Qwen3.5-35B-A3B 65.3%
- 6
Qwen 3.8 27B 63.5%
- 7
Gemini 3 Pro 63.4%
- 8
Qwen3-Omni-30B-A3B 61.3%
- 9
Nemotron Nano 12B v2 VL 61.2%
- 10 GLM-4.6V-Flash 59%
- 11
Qwen3.5-9B 58.7%
- 12
Muse Glimmer 56.9%
- 13
Llama-3.1-Nemotron-Nano-VL-8B-V1 56.4%
- 14 GPT-5 55.5%
- 15 LFM2.5-VL-3B 42.6%
Source: Official OCRBench v2 leaderboard
- 74.2
Qwen 3.8 Max
- 70.7
Qwen3.7-Plus
- 68.5
MiniCPM-V-4.7-35B-A3B
- 68.3
Qwen3.6 27B
- 68.1
Nemotron 3.5 Super VL
- 67.0
Nemotron 3 Nano Omni 30B-A3B
- 66.8
Qwen3-VL-235B-A22B
- 65.5
Qwen 3.6 35B-A3B
- 65.3
Qwen3.5-35B-A3B
- 65.1 GLM-4.6V
- 64.5
Nova 2.0 Pro Preview
- 63.5
Qwen 3.8 27B
- 63.4
Gemini 3 Pro
- 63.2
EXAONE 4.5 33B
- 63.2
Seed2.1 Pro
- 62.5
Seed2.0 Pro
- 62.0
Nemotron Nano 12B v2 VL
- 61.5
Qwen2.5-VL-72B
- 61.3
Qwen3-Omni-30B-A3B
- 60.1
Llama-3.1-Nemotron-Nano-VL-8B-V1
- 60.1
Intern-S1-Pro
- 60.1
Cosmos3-Nano
- 59.0 GLM-4.6V-Flash
- 58.7
Qwen3.5-9B
- 58.2
Nova 2 Omni Preview
- 57.8
Qwen2.5-Omni-7B
- 56.9
Amazon Nova Premier
- 56.9
Muse Glimmer
- 56.1
Nova 2 Lite
- 55.5 GPT-5
- 54.7
MiniCPM-OCR-1.0
- 47.5 LFM2.5-VL-3B
- 43.7
Cosmos3-Edge
- 36.7
North Micro Vision Instruct
- 1
Qwen 3.8 Max 74.2%
- 2
Qwen3.7-Plus 70.7%
- 3
MiniCPM-V-4.7-35B-A3B 68.5%
- 4
Qwen3.6 27B 68.3%
- 5
Nemotron 3.5 Super VL 68.1%
- 6
Nemotron 3 Nano Omni 30B-A3B 67%
- 7
Qwen3-VL-235B-A22B 66.8%
- 8
Qwen 3.6 35B-A3B 65.5%
- 9
Qwen3.5-35B-A3B 65.3%
- 10 GLM-4.6V 65.1%
- 11
Nova 2.0 Pro Preview 64.5%
- 12
Qwen 3.8 27B 63.5%
- 13
Gemini 3 Pro 63.4%
- 14
EXAONE 4.5 33B 63.2%
- 15
Seed2.1 Pro 63.2%
- 16
Seed2.0 Pro 62.5%
- 17
Nemotron Nano 12B v2 VL 62%
- 18
Qwen2.5-VL-72B 61.5%
- 19
Qwen3-Omni-30B-A3B 61.3%
- 20
Llama-3.1-Nemotron-Nano-VL-8B-V1 60.1%
- 21
Intern-S1-Pro 60.1%
- 22
Cosmos3-Nano 60.1%
- 23 GLM-4.6V-Flash 59%
- 24
Qwen3.5-9B 58.7%
- 25
Nova 2 Omni Preview 58.2%
- 26
Qwen2.5-Omni-7B 57.8%
- 27
Amazon Nova Premier 56.9%
- 28
Muse Glimmer 56.9%
- 29
Nova 2 Lite 56.1%
- 30 GPT-5 55.5%
- 31
MiniCPM-OCR-1.0 54.7%
- 32 LFM2.5-VL-3B 47.5%
- 33
Cosmos3-Edge 43.7%
- 34
North Micro Vision Instruct 36.7%
Sources: Labs’ model cards and official OCRBench v2 leaderboard
| # | Entry | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | KDL Frontier (2026-10, EN) | 76.9% | Sep 2026 | Official leaderboard | ||
| 2 | SEEU DOC | 76.2% | Sep 2026 | Official leaderboard | ||
| 3 | Yaochi-DTS-A2T-1.9 | 73.4% | Sep 2026 | Official leaderboard | ||
| 4 | U2-OCR | 71.7% | Sep 2026 | Official leaderboard | ||
| 5 | Posicube_DVLM | 69.8% | Sep 2026 | Official leaderboard | ||
| 6 | | 68.3% | Apr 2026 | Sep 2026 | Official leaderboard | |
| 7 | KDL Frontier | 68.1% | Sep 2026 | Official leaderboard | ||
| 8 | | 67.4% | Oct 2026 | Sep 2026 | Official leaderboard | |
| 9 | | 65.8% | Apr 2026 | Sep 2026 | Official leaderboard | |
| 10 | | 65.5% | Apr 2026 | Sep 2026 | Official leaderboard | |
| 11 | | 65.3% | Feb 2026 | Sep 2026 | Official leaderboard | |
| 12 | | 63.5% | Aug 2026 | Sep 2026 | Official leaderboard | |
| 13 | | 63.4% | Nov 2025 | Sep 2026 | Official leaderboard | |
| 14 | Seed1.6-vision | 62.2% | Sep 2026 | Official leaderboard | ||
| 15 | TeleMM-2.0 | 61.8% | Sep 2026 | Official leaderboard | ||
| 16 | | 61.3% | Sep 2025 | Sep 2026 | Official leaderboard | |
| 17 | | 61.2% | Oct 2025 | Sep 2026 | Official leaderboard | |
| 18 | SenseNova-U1.5-8B-MoT | 59.9% | Sep 2026 | Official leaderboard | ||
| 19 | Gemini-2.5-Pro (snapshot unspecified) | 59.3% | Sep 2026 | Official leaderboard | ||
| 20 | GLM-4.6V-Flash | 59% | Dec 2025 | Sep 2026 | Official leaderboard | |
| 21 | | 58.7% | Feb 2026 | Sep 2026 | Official leaderboard | |
| 22 | SenseNova-U1 | 58.2% | Sep 2026 | Official leaderboard | ||
| 23 | | 56.9% | Aug 2026 | Sep 2026 | Official leaderboard | |
| 24 | | 56.4% | Jun 2025 | Sep 2026 | Official leaderboard | |
| 25 | GPT-5 | 55.5% | Aug 2025 | Sep 2026 | Official leaderboard | |
| 26 | LFM2.5-VL-3B | 42.6% | Aug 2026 | Sep 2026 | Official leaderboard |
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | | 68.3% | Apr 2026 | Sep 2026 | Official leaderboard | |
| 2 | | 67.4% | Oct 2026 | Sep 2026 | Official leaderboard | |
| 3 | | 65.8% | Apr 2026 | Sep 2026 | Official leaderboard | |
| 4 | | 65.5% | Apr 2026 | Sep 2026 | Official leaderboard | |
| 5 | | 65.3% | Feb 2026 | Sep 2026 | Official leaderboard | |
| 6 | | 63.5% | Aug 2026 | Sep 2026 | Official leaderboard | |
| 7 | | 63.4% | Nov 2025 | Sep 2026 | Official leaderboard | |
| 8 | | 61.3% | Sep 2025 | Sep 2026 | Official leaderboard | |
| 9 | | 61.2% | Oct 2025 | Sep 2026 | Official leaderboard | |
| 10 | GLM-4.6V-Flash | 59% | Dec 2025 | Sep 2026 | Official leaderboard | |
| 11 | | 58.7% | Feb 2026 | Sep 2026 | Official leaderboard | |
| 12 | | 56.9% | Aug 2026 | Sep 2026 | Official leaderboard | |
| 13 | | 56.4% | Jun 2025 | Sep 2026 | Official leaderboard | |
| 14 | GPT-5 | 55.5% | Aug 2025 | Sep 2026 | Official leaderboard | |
| 15 | LFM2.5-VL-3B | 42.6% | Aug 2026 | Sep 2026 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | 74.2% | Aug 2026 | Lab model card | ||
| 2 | 70.7% | Jun 2026 | Lab model card | ||
| 3 | 68.5% | Oct 2026 | Lab model card | ||
| 4 | 68.3% | Apr 2026 | Official leaderboard | ||
| 5 | 68.1% | Oct 2026 | Lab model card | ||
| 6 | 67% | Apr 2026 | Lab model card | ||
| 7 | 66.8% | Sep 2025 | Lab model card | ||
| 8 | 65.5% | Apr 2026 | Official leaderboard | ||
| 9 | 65.3% | Feb 2026 | Official leaderboard | ||
| 10 | GLM-4.6V | 65.1% | Dec 2025 | Lab model card | |
| 11 | 64.5% | Dec 2025 | Lab model card | ||
| 12 | 63.5% | Aug 2026 | Official leaderboard | ||
| 13 | 63.4% | Nov 2025 | Official leaderboard | ||
| 14 | 63.2% | Apr 2026 | Lab model card | ||
| 15 | 63.2% | Jun 2026 | Lab model card | ||
| 16 | 62.5% | Feb 2026 | Lab model card | ||
| 17 | 62% | Oct 2025 | Lab model card | ||
| 18 | 61.5% | Jan 2025 | Lab model card | ||
| 19 | 61.3% | Sep 2025 | Official leaderboard | ||
| 20 | 60.1% | Jun 2025 | Lab model card | ||
| 21 | 60.1% | Feb 2026 | Lab model card | ||
| 22 | 60.1% | May 2026 | Lab model card | ||
| 23 | GLM-4.6V-Flash | 59% | Dec 2025 | Official leaderboard | |
| 24 | 58.7% | Feb 2026 | Official leaderboard | ||
| 25 | 58.2% | Dec 2025 | Lab model card | ||
| 26 | 57.8% | Mar 2025 | Lab model card | ||
| 27 | 56.9% | Apr 2025 | Lab model card | ||
| 28 | 56.9% | Aug 2026 | Official leaderboard | ||
| 29 | 56.1% | Dec 2025 | Lab model card | ||
| 30 | GPT-5 | 55.5% | Aug 2025 | Official leaderboard | |
| 31 | 54.7% | Oct 2026 | Lab model card | ||
| 32 | LFM2.5-VL-3B | 47.5% | Aug 2026 | Lab model card | |
| 33 | 43.7% | May 2026 | Lab model card | ||
| 34 | 36.7% | Aug 2026 | Lab model card |
Labs report scores for 24 models in their cards. Switch the scores above to rank them together.
Rankings by part
We read 9 parts from the benchmark’s own leaderboard in 2 groups: task type and language. Each ranks models on one part of the test. Gemini 3 Pro leads 3 of them, and 5 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Recognition
15 models we track have a score here.
- 1
Qwen 3.8 27B ↑5 75.5%
- 2 GLM-4.6V-Flash ↑8 75.3%
- 3
Qwen 3.6 35B-A3B ↑1 74.9%
- 4
Nemotron 3.5 Super VL ↓2 74%
- 5
Qwen3.5-9B ↑6 72.4%
- 6
Qwen3-Omni-30B-A3B ↑2 72.3%
- 7
Qwen3.6 27B ↓6 71.1%
- 8
Nemotron 3 Nano Omni 30B-A3B ↓5 70.5%
- 9
Qwen3.5-35B-A3B ↓4 69.8%
- 10 GPT-5 ↑4 69.3%
Referring
15 models we track have a score here.
- 1
Gemini 3 Pro ↑6 65.9%
- 2
Qwen3.5-35B-A3B ↑3 65.2%
- 3
Qwen 3.6 35B-A3B ↑1 62.8%
- 4
Qwen3-Omni-30B-A3B ↑4 62%
- 5
Llama-3.1-Nemotron-Nano-VL-8B-V1 ↑8 61.3%
- 6
Nemotron 3.5 Super VL ↓4 61%
- 7
Qwen3.6 27B ↓6 58.4%
- 8
Nemotron 3 Nano Omni 30B-A3B ↓5 56.2%
- 9
Qwen 3.8 27B ↓3 55.6%
- 10
Nemotron Nano 12B v2 VL ↓1 54.5%
Spotting
15 models we track have a score here.
- 1
Gemini 3 Pro ↑6 80.8%
- 2
Qwen3.6 27B ↓1 74.3%
- 3
Llama-3.1-Nemotron-Nano-VL-8B-V1 ↑10 68.6%
- 4
Nemotron 3.5 Super VL ↓2 63.2%
- 5
Qwen 3.8 27B ↑1 61.7%
- 6
Qwen 3.6 35B-A3B ↓2 58.9%
- 7
Qwen3.5-35B-A3B ↓2 58.6%
- 8
Nemotron 3 Nano Omni 30B-A3B ↓5 54.2%
- 9
Qwen3-Omni-30B-A3B ↓1 45.6%
- 10
Qwen3.5-9B ↑1 41.8%
Extraction
15 models we track have a score here.
- 1
Muse Glimmer ↑11 94.3%
- 2
Gemini 3 Pro ↑5 93.9%
- 3
Qwen3-Omni-30B-A3B ↑5 93.5%
- 4
Nemotron 3 Nano Omni 30B-A3B ↓1 93.3%
- 5
Qwen3.6 27B ↓4 92.8%
- 6
Nemotron 3.5 Super VL ↓4 92.2%
- 7
Nemotron Nano 12B v2 VL ↑2 92%
- 8
Qwen 3.6 35B-A3B ↓4 91.5%
- 9
Qwen3.5-9B ↑2 91.1%
- 10
Qwen3.5-35B-A3B ↓5 90.5%
Parsing
15 models we track have a score here.
- 1
Gemini 3 Pro ↑6 27.1%
- 2
Nemotron Nano 12B v2 VL ↑7 26.6%
- 3
Qwen3.6 27B ↓2 26.4%
- 4
Nemotron 3 Nano Omni 30B-A3B ↓1 26.3%
- 5
Nemotron 3.5 Super VL ↓3 26.3%
- 6
Qwen3.5-35B-A3B ↓1 24.9%
- 7 GPT-5 ↑7 24.8%
- 8
Qwen 3.8 27B ↓2 24.8%
- 9
Muse Glimmer ↑3 24.5%
- 10
Qwen 3.6 35B-A3B ↓6 23.2%
Calculation
15 models we track have a score here.
- 1 GLM-4.6V-Flash ↑9 90.4%
- 2
Nemotron 3.5 Super VL 87.9%
- 3
Nemotron 3 Nano Omni 30B-A3B 86.9%
- 4 GPT-5 ↑10 83.7%
- 5
Qwen3.6 27B ↓4 81.9%
- 6
Nemotron Nano 12B v2 VL ↑3 80.4%
- 7
Qwen 3.6 35B-A3B ↓3 78%
- 8
Muse Glimmer ↑4 76.9%
- 9
Qwen3.5-35B-A3B ↓4 74.4%
- 10
Qwen 3.8 27B ↓4 67.9%
Understanding
15 models we track have a score here.
- 1
Qwen3.5-35B-A3B ↑4 84.1%
- 2
Qwen3.6 27B ↓1 83.5%
- 3
Nemotron 3 Nano Omni 30B-A3B 82%
- 4
Qwen 3.8 27B ↑2 81.1%
- 5
Nemotron 3.5 Super VL ↓3 80.5%
- 6
Qwen 3.6 35B-A3B ↓2 80.2%
- 7 GLM-4.6V-Flash ↑3 79.8%
- 8
Qwen3.5-9B ↑3 79.7%
- 9
Gemini 3 Pro ↓2 77.8%
- 10 GPT-5 ↑4 76.4%
Reasoning
15 models we track have a score here.
- 1
Muse Glimmer ↑11 67%
- 2 GPT-5 ↑12 65.4%
- 3 GLM-4.6V-Flash ↑7 59%
- 4
Qwen3.6 27B ↓3 57.9%
- 5
Nemotron 3 Nano Omni 30B-A3B ↓2 57.1%
- 6
Nemotron Nano 12B v2 VL ↑3 57%
- 7
Qwen3-Omni-30B-A3B ↑1 55.3%
- 8
Qwen3.5-35B-A3B ↓3 55.3%
- 9
Qwen 3.6 35B-A3B ↓5 54.8%
- 10
Nemotron 3.5 Super VL ↓8 54%
Chinese
12 models we track have a score here.
- 1
Qwen3.5-9B ↑10 64.1%
- 2
Gemini 3 Pro ↑5 63.8%
- 3
Qwen3.5-35B-A3B ↑2 61.7%
- 4
Qwen 3.6 35B-A3B 60.8%
- 5
Qwen3-Omni-30B-A3B ↑3 60%
- 6 GLM-4.6V-Flash ↑4 59.5%
- 7
Nemotron 3 Nano Omni 30B-A3B ↓4 59%
- 8
Qwen3.6 27B ↓7 58.7%
- 9
Qwen 3.8 27B ↓3 56.9%
- 10
Muse Glimmer ↑2 56.8%
6 models have held number one on the official leaderboard since its release in December 2024 5 models have held number one with claims included since its release in December 2024
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
Sources: Official OCRBench v2 leaderboard and labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official OCRBench v2 leaderboard and labs’ model cards
5 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Oct 28, 2026 | 54.7 | OCRBench v2 | |
| Oct 15, 2026 | 68.1 | OCRBench v2, EN 68.12 / ZH 59.15 | |
| Oct 7, 2026 | 68.5 | OCRBench_v2 | |
| Aug 12, 2026 | LFM2.5-VL-3B | 47.5 | OCRBench v2 (En) |
| Aug 12, 2026 | 36.7 | OCRBench v2_en | |
| Aug 3, 2026 | 74.2 | OCR-Bench-V2 (EN/ZH) 74.2 / 68.3 | |
| Jun 1, 2026 | 70.7 | OCR-Bench-V2(EN) 70.7, OCR-Bench-V2(ZH) 67.1 | |
| Apr 28, 2026 | 67 | OCRBenchV2 (EN) | |
| Apr 9, 2026 | 63.2 | OCRBench v2 | |
| Feb 2, 2026 | 60.1 | OCRBench V2 (ENG / CHN) 60.1 / 60.6 | |
| Dec 8, 2025 | GLM-4.6V | 65.1 | OCR-Bench_v2 (EN) |
| Dec 2, 2025 | 58.2 | OCRBench-v2 avg accuracy | |
| Dec 2, 2025 | 56.1 | OCRBench-v2 avg accuracy | |
| Dec 2, 2025 | 64.5 | OCRBench-v2 avg accuracy | |
| Oct 28, 2025 | 62 | OCRBenchv2 (English) | |
| Sep 23, 2025 | 66.8 | OCRBenchV2 (en/ch) 66.8 / 63.5 | |
| Jun 3, 2025 | 60.1 | OCRBenchV2 English 60.1 / Chinese 37.9 | |
| Apr 30, 2025 | 56.9 | OCRBench-v2 | |
| Mar 27, 2025 | 57.8 | OCRBench_V2en | |
| Jan 28, 2025 | 61.5 | OCRBench-V2(en/zh) 61.5/63.7 |
16 system cards report it, from 6 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Oct 15, 2026 | | 68.1 | OCRBench-v2 (EN) 68.12, (ZH) 59.15 (Table 10) | Capabilities |
| Jun 23, 2026 | | 63.2 | OCRBenchv2 (Table 9) | Capabilities |
| May 31, 2026 | | 60.1 | OCRBench-v2, Cosmos3-Nano 8B column, Table 10 | Capabilities |
| May 31, 2026 | | 43.7 | OCRBench-v2, Cosmos3-Edge 2B column, Table 10 | Capabilities |
| Apr 27, 2026 | | 67 | OCRBenchV2 (EN), reasoning on, ZH 52.7. Reasoning off EN 65.8, ZH 52.0 (Table 7) | Capabilities |
| Apr 9, 2026 | | 63.2 | OCRBench v2 (Table 2, Reasoning mode) | Capabilities |
| Mar 26, 2026 | | 60.1 | OCRBench V2 (ENG / CHN) 60.1 / 60.6, English recorded, non-thinking (Table 2) | Capabilities |
| Feb 14, 2026 | | 62.5 | OCRBenchv2, mean of English and Chinese overall (Table 8) | Capabilities |
| Dec 2, 2025 | | 58.2 | OCRBench-v2 avg accuracy (Table 6). The 1.2 text says 58.5 | Capabilities |
| Dec 2, 2025 | | 56.1 | OCRBench-v2 avg accuracy (Table 3) | Capabilities |
| Dec 2, 2025 | | 64.5 | OCRBench-v2 avg accuracy (Table 3) | Capabilities |
| Nov 26, 2025 | | OCRBench_v2 en 66.8, zh 63.5, no overall (Table 2) | Capabilities | |
| Nov 6, 2025 | | 54.8 | OCRBenchV2 EN, reasoning on, ZH 39.8 (Table 3). Reasoning off EN 62.0, ZH 44.2 | Capabilities |
| Apr 30, 2025 | | 56.9 | OCRBench-v2 (Table 1) | Capabilities |
| Mar 26, 2025 | | 57.8 | OCRBench_V2 en (Table 5) | Capabilities |
| Feb 19, 2025 | | OCRBench_v2 en 61.5, zh 63.7, no overall (Table 5) | Capabilities |
The lab's number comes out ahead in 5 of 5 pairs that differ
Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. Every benchmark.
| Model | Lab says | Board says | Gap |
|---|---|---|---|
| LFM2.5-VL-3B | 47.5% OCRBench v2 (En) | 42.6% Leaderboard | +4.9 pts |
| | 60.1% OCRBenchV2 English 60.1 / Chinese 37.9 | 56.4% Leaderboard | +3.7 pts |
| | 67% OCRBenchV2 (EN) | 65.8% Leaderboard | +1.2 pts |
| | 62% OCRBenchv2 (English) | 61.2% Leaderboard | +0.8 pts |
| | 68.1% OCRBench v2, EN 68.12 / ZH 59.15 | 67.4% Leaderboard | +0.7 pts |
2 logged changes that move scores
Changes by the maintainers, so a score from before a change may not compare with one from after it. Newest first, each with the maintainers' own source. The list may be incomplete. What we log.
- Grader changed July 5th, 2025
The maintainers corrected English answer-length routing. Longer answers now receive edit-distance scoring, while short answers use substring matching. Source
- Tasks changed June 21st, 2025
The maintainers introduced a private evaluation dataset and quarterly leaderboard updates. Private-set results measure a different question set from the public release. Source
Versions
We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.
Sources: Official leaderboards and labs’ model cards
More in Vision
- MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy.
Qwen 3.6 Plus 86%
- MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
- CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
- MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy.
Seed2.1 Pro 90.7%
- MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy.
Qwen 3.8 Max 97.7%
- MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy.
Qwen 3.8 Max 85.9%
Data audit
All 11 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access Public tasks, plus a private test set.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (26 entries), launch posts (20 scores) and system cards (14 scores).
- Pass Models scored 34 models we track, plus 11 other entries on its leaderboard.
- Pass Newest score The newest is from October 2026, for MiniCPM-OCR-1.0.
- Note Frontier models None of the top 10 has a score. The strongest tested is Qwen 3.8 Max (#30).
- Pass Headroom The best is 74.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on 99franklin.github.io: 15 rows for models we track and 11 other entries at the last read.
- Pass Last read November 18th, 2026.
- Pass Read worked The last read worked.
Consistency 2 pass
- Pass Lab claims All 5 pairs agree within 10 points.
- Pass Known issues No published audit found problems.