Benchmark Almanac

Extract Bench (Contextual AI)

Gemini 3 Flash 0.07 leads since February 2026
Not in model cards

Models extract structured JSON from PDFs. Field checks include exact matches, numerical tolerances and semantic judgments.

The suite has 35 tasks across five schemas.

Made by
Contextual AI
Released
February 2026
Kind
Documents, charts and text in images
Level
University
Scored
Graded by a model
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Documents, Text → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-5.2 #51 on the capability index
Last model added
December 2025
Scores refreshed
November 20th, 2026
6
Models scored in its paper

Gemini 3 Flash leads with 0.07 among models released by December 2025

Best score per model from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

The newest model with a score here came out in December 2025, 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 Flash Baseline 0.07
  2. 2 Gemini 3 Pro Baseline 0.06
  3. 3 GPT-5.2 Structured outputs 0.05
  4. 4 GPT-5 Baseline 0.04
  5. 5 Claude Sonnet 4.5 Baseline 0.03
  6. 6 Claude Opus 4.5 Baseline 0.03

Source: The Extract Bench (Contextual AI) paper

#ModelScoreSettingDateSource
1 Gemini 3 Flash 0.07 Baseline Dec 2025 Official leaderboard
2 Gemini 3 Pro 0.06 Baseline Nov 2025 Official leaderboard
3 GPT-5.2 0.05 Structured outputs Dec 2025 Official leaderboard
4 GPT-5 0.04 Baseline Aug 2025 Official leaderboard
5 Claude Sonnet 4.5 0.03 Baseline Sep 2025 Official leaderboard
6 Claude Opus 4.5 0.03 Baseline Nov 2025 Official leaderboard

Rankings by part

We read 5 parts from the benchmark’s paper, each ranking models on one part of the test, such as Credit, Research and Resumes. Gemini 3 Flash leads 2 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Credit

6 models we track have a score here.

  1. 1 Gemini 3 Pro ↑1 0.89
  2. 2 GPT-5 ↑2 0.87
  3. 3 Gemini 3 Flash ↓2 0.85
  4. 4 GPT-5.2 ↓1 0.83
  5. 5 Claude Sonnet 4.5 0.00
  6. 6 Claude Opus 4.5 0.00

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Gemini 3 Flash
Structured … Baseline
0.07
Gemini 3 Pro
Structured … Baseline
0.06
GPT-5.2
Baseline
0.05
GPT-5
Baseline
0.04
Claude Sonnet 4.5
Structured … Baseline
0.03
Claude Opus 4.5
Structured … Baseline
0.03

Source: The Extract Bench (Contextual AI) paper

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3 Flash Baseline 0.07
Gemini 3 Pro Baseline 0.06
Gemini 3 Flash Structured outputs 0.06
GPT-5.2 Structured outputs 0.05
GPT-5.2 Baseline 0.05
Gemini 3 Pro Structured outputs 0.04
GPT-5 Baseline 0.04
GPT-5 Structured outputs 0.04
Claude Sonnet 4.5 Baseline 0.03
Claude Opus 4.5 Baseline 0.03
Claude Sonnet 4.5 Structured outputs 0.00
Claude Opus 4.5 Structured outputs 0.00

Number one over time

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

Google DeepMind
Benchmark paper Not tested since Feb 2026

Source: The Extract Bench (Contextual AI) paper

Scores over time

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

Plot with other benchmarks
In the benchmark’s paper
0.00 0.02 0.04 0.06 0.08 2026 0.00 0.02 0.04 0.06 0.08 2026

Source: Official Extract Bench (Contextual AI) 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

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 11 months ago, in December 2025, for Gemini 3 Flash. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the benchmark’s paper (6 entries).
  • Pass Models scored 6 models we track.
  • Warning Newest score The newest came 11 months ago, in December 2025, for Gemini 3 Flash.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.2 (#51).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 12 rows for models we track 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