Benchmark Almanac

IDP Core Bench

Gemini 3.1 Pro 89.6% leads since February 2026
Not in model cards

Models extract fields and tables, transcribe text, and answer questions about documents. The test covers about 2,000 invoices, receipts, forms, and handwritten documents.

The makers list 27 models on the four-task leaderboard. They publish an overall score and a score for each task.

Made by
Nanonets and IIT Indore
Kind
Documents, charts and text in images
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
In and out
Documents, Text → Text, Tables
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-5.4 #28 on the capability index
Last model added
March 2026
Scores refreshed
November 20th, 2026
27
Entries on its own leaderboard, 23 models we track

Gemini 3.1 Pro leads with 89.6% among models released by March 2026 Gemini 3.1 Pro leads with 89.6% among models released by March 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 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. This view shows only models we track, each at its best.

The newest model with a score here came out in March 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Gemini 3.1 Pro 89.6%
  2. 2 GPT-5.4 84.4%
  3. 3 Gemini 3 Pro 81.8%
  4. 4 Claude Sonnet 4.6 81.2%
  5. 5 Claude Opus 4.6 81.1%
  6. 6 Gemini 3 Flash 80.5%
  7. 7 Nanonets OCR-3 80.2%
  8. 8 Qwen3-VL-235B 80%
  9. 9 Qwen3-VL-Plus 79.8%
  10. 10 GPT-5.2 77.4%
  11. 11 Qwen3.5-9B 76.2%
  12. 12 GPT-4.1 74.7%
  13. 13 Qwen3.5-4B 74.5%
  14. 14 Nanonets OCR2+ 73.8%
  15. 15 GPT-5 mini 73.3%
  16. 16 Claude Haiku 4.5 72.9%
  17. 17 Ministral 8B 71.7%
  18. 18 Mistral Small 4 68.5%
  19. 19 Qwen3.5-2B 67.1%
  20. 20 GPT-5 nano 65.8%
  21. 21 Qwen3.5-0.8B 61.2%
  22. 22 Pixtral 12B 59%
  23. 23 Llama-3.2-Vision-11B 58.6%
  24. 24 Gemma 4 E4B 55%
  25. 25 GLM-OCR 54.9%
  26. 26 Gemma 4 E2B 44.1%
  27. 27 Gemma 3 12B 0%

Source: Official IDP Core Bench leaderboard

#EntryScoreSettingDateLinkSource
1 Gemini 3.1 Pro 89.6% Feb 2026 idp-leaderboard.org Official leaderboard
2 GPT-5.4 84.4% Mar 2026 idp-leaderboard.org Official leaderboard
3 Gemini 3 Pro 81.8% Nov 2025 idp-leaderboard.org Official leaderboard
4 Claude Sonnet 4.6 81.2% Feb 2026 idp-leaderboard.org Official leaderboard
5 Claude Opus 4.6 81.1% Feb 2026 idp-leaderboard.org Official leaderboard
6 Gemini 3 Flash 80.5% Dec 2025 idp-leaderboard.org Official leaderboard
7 Nanonets OCR-3 closed weights 80.2% idp-leaderboard.org Official leaderboard
8 Qwen3-VL-235B open weights 80% idp-leaderboard.org Official leaderboard
9 Qwen3-VL-Plus 79.8% Sep 2025 idp-leaderboard.org Official leaderboard
10 GPT-5.2 77.4% Dec 2025 idp-leaderboard.org Official leaderboard
11 Qwen3.5-9B 76.2% Feb 2026 idp-leaderboard.org Official leaderboard
12 GPT-4.1 74.7% Apr 2025 idp-leaderboard.org Official leaderboard
13 Qwen3.5-4B 74.5% Feb 2026 idp-leaderboard.org Official leaderboard
14 Nanonets OCR2+ closed weights 73.8% idp-leaderboard.org Official leaderboard
15 GPT-5 mini 73.3% Aug 2025 idp-leaderboard.org Official leaderboard
16 Claude Haiku 4.5 72.9% Oct 2025 idp-leaderboard.org Official leaderboard
17 Ministral 8B 71.7% Oct 2024 idp-leaderboard.org Official leaderboard
18 Mistral Small 4 68.5% Mar 2026 idp-leaderboard.org Official leaderboard
19 Qwen3.5-2B 67.1% Feb 2026 idp-leaderboard.org Official leaderboard
20 GPT-5 nano 65.8% Aug 2025 idp-leaderboard.org Official leaderboard
21 Qwen3.5-0.8B 61.2% Feb 2026 idp-leaderboard.org Official leaderboard
22 Pixtral 12B 59% Sep 2024 idp-leaderboard.org Official leaderboard
23 Llama-3.2-Vision-11B open weights 58.6% idp-leaderboard.org Official leaderboard
24 Gemma 4 E4B 55% Mar 2026 idp-leaderboard.org Official leaderboard
25 GLM-OCR 54.9% Jan 2026 idp-leaderboard.org Official leaderboard
26 Gemma 4 E2B 44.1% Mar 2026 idp-leaderboard.org Official leaderboard
27 Gemma 3 12B 0% Mar 2025 idp-leaderboard.org Official leaderboard
#ModelScoreSettingDateLinkSource
1 Gemini 3.1 Pro 89.6% Feb 2026 idp-leaderboard.org Official leaderboard
2 GPT-5.4 84.4% Mar 2026 idp-leaderboard.org Official leaderboard
3 Gemini 3 Pro 81.8% Nov 2025 idp-leaderboard.org Official leaderboard
4 Claude Sonnet 4.6 81.2% Feb 2026 idp-leaderboard.org Official leaderboard
5 Claude Opus 4.6 81.1% Feb 2026 idp-leaderboard.org Official leaderboard
6 Gemini 3 Flash 80.5% Dec 2025 idp-leaderboard.org Official leaderboard
7 Qwen3-VL-Plus 79.8% Sep 2025 idp-leaderboard.org Official leaderboard
8 GPT-5.2 77.4% Dec 2025 idp-leaderboard.org Official leaderboard
9 Qwen3.5-9B 76.2% Feb 2026 idp-leaderboard.org Official leaderboard
10 GPT-4.1 74.7% Apr 2025 idp-leaderboard.org Official leaderboard
11 Qwen3.5-4B 74.5% Feb 2026 idp-leaderboard.org Official leaderboard
12 GPT-5 mini 73.3% Aug 2025 idp-leaderboard.org Official leaderboard
13 Claude Haiku 4.5 72.9% Oct 2025 idp-leaderboard.org Official leaderboard
14 Ministral 8B 71.7% Oct 2024 idp-leaderboard.org Official leaderboard
15 Mistral Small 4 68.5% Mar 2026 idp-leaderboard.org Official leaderboard
16 Qwen3.5-2B 67.1% Feb 2026 idp-leaderboard.org Official leaderboard
17 GPT-5 nano 65.8% Aug 2025 idp-leaderboard.org Official leaderboard
18 Qwen3.5-0.8B 61.2% Feb 2026 idp-leaderboard.org Official leaderboard
19 Pixtral 12B 59% Sep 2024 idp-leaderboard.org Official leaderboard
20 Gemma 4 E4B 55% Mar 2026 idp-leaderboard.org Official leaderboard
21 GLM-OCR 54.9% Jan 2026 idp-leaderboard.org Official leaderboard
22 Gemma 4 E2B 44.1% Mar 2026 idp-leaderboard.org Official leaderboard
23 Gemma 3 12B 0% Mar 2025 idp-leaderboard.org Official leaderboard

Rankings by part

We read 4 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as KIE, OCR and Table. Gemini 3.1 Pro leads 3 of them, and Gemini 3 Flash leads the other one. Arrows show how far a model moved from its overall rank among the models we track.

KIE

23 models we track have a score here.

  1. 1 Gemini 3 Flash ↑5 91.1%
  2. 2 Claude Opus 4.6 ↑3 89.8%
  3. 3 Claude Sonnet 4.6 ↑1 89.5%
  4. 4 GPT-5.2 ↑4 87.5%
  5. 5 GPT-4.1 ↑5 87.1%
  6. 6 Gemini 3.1 Pro ↓5 86.8%
  7. 7 Qwen3.5-9B ↑2 86.5%
  8. 8 Qwen3.5-4B ↑3 86%
  9. 9 Ministral 8B ↑5 85.7%
  10. 10 GPT-5 mini ↑2 85.7%

6 models have held number one on the official leaderboard since 2023

Lab claims count from the day the card came out, independent runs from the model’s release.

Mistral AI OpenAI Qwen Google DeepMind
Own leaderboard Passed 85% in Feb 2026, by Gemini 3.1 Pro

Source: Official IDP Core Bench leaderboard

Gemini 3.1 Pro on the official leaderboard passed 85% in February 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemini 3.1 Pro Gemini 3 Pro GPT-4.1 Ministral 8B Pixtral 12B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemini 3.1 Pro Gemini 3 Pro

Source: Official IDP Core Bench leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Vision

All 636 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 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the official leaderboard (27 entries).
  • Pass Models scored 23 models we track, plus 4 other entries on its leaderboard.
  • Note Newest score The newest came 8 months ago, in March 2026, for Gemma 4 E2B. That’s expected for a saturated test.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is GPT-5.4 (#28).
  • Note Headroom Saturated. The best is 89.6%, past 85% since February 2026.
Leaderboard import 3 pass
  • Pass Importer We read the board on idp-leaderboard.org: 23 rows for models we track and 4 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check