Benchmark Almanac

Multimodal RewardBench 2

Gemini 3 Pro 76.3% leads since December 2025
Not in model cards
Version 2

Reward models compare two outputs containing text and images. Accuracy measures agreement with expert preferences across four tasks.

The evaluation contains 4,000 labeled pairs.

Made by
Meta and Surge AI
Released
December 2025
Kind
General visual questions
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Preference
Language
English
In and out
Text, Images → Text
Interaction
Single turn
Run by
Anyone
In views
Vision
Strongest model tested
Gemini 3 Pro #56 on the capability index
Last model added
November 2025
Scores refreshed
November 20th, 2026
15
Entries on its own leaderboard, 12 models we track

Gemini 3 Pro leads with 76.3% among models released by November 2025 Gemini 3 Pro leads with 76.3% among models released by November 2025

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 November 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 Pro 76.3%
  2. 2 GPT-5 72.2%
  3. 3 Gemini 2.5 Pro Jun 2025 70.9%
  4. 4 Qwen3-VL-32B 64.6%
  5. 5 Gemini 2.5 Flash Sep 2025 64.1%
  6. 6 GPT-4.1 63.5%
  7. 7 Qwen3-VL-235B-A22B 62.9%
  8. 8 GPT-4o Mar 2025 59.7%
  9. 9 Qwen3-VL-8B 59.3%
  10. 10 Qwen2.5-VL-72B 59%
  11. 11 Qwen3-VL-30BA3B 58.5%
  12. 12 Gemma 3 27B 57.3%
  13. 13 Gemma 3 12B 55.3%
  14. 14 Qwen2.5-VL-7B 50.9%
  15. 15 Gemma 3 4B 50.7%

Source: Official Multimodal RewardBench 2 leaderboard

#EntryScoreSettingDateSource
1 Gemini 3 Pro 76.3% Nov 2025 Official leaderboard
2 GPT-5 72.2% Aug 2025 Official leaderboard
3 Gemini 2.5 Pro Jun 2025 70.9% Jun 2025 Official leaderboard
4 Qwen3-VL-32B open weights 64.6% Official leaderboard
5 Gemini 2.5 Flash Sep 2025 64.1% Sep 2025 Official leaderboard
6 GPT-4.1 63.5% Apr 2025 Official leaderboard
7 Qwen3-VL-235B-A22B 62.9% Sep 2025 Official leaderboard
8 GPT-4o Mar 2025 59.7% Mar 2025 Official leaderboard
9 Qwen3-VL-8B open weights 59.3% Official leaderboard
10 Qwen2.5-VL-72B 59% Jan 2025 Official leaderboard
11 Qwen3-VL-30BA3B open weights 58.5% Official leaderboard
12 Gemma 3 27B 57.3% Mar 2025 Official leaderboard
13 Gemma 3 12B 55.3% Mar 2025 Official leaderboard
14 Qwen2.5-VL-7B 50.9% Jan 2025 Official leaderboard
15 Gemma 3 4B 50.7% Mar 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 3 Pro 76.3% Nov 2025 Official leaderboard
2 GPT-5 72.2% Aug 2025 Official leaderboard
3 Gemini 2.5 Pro Jun 2025 70.9% Jun 2025 Official leaderboard
4 Gemini 2.5 Flash Sep 2025 64.1% Sep 2025 Official leaderboard
5 GPT-4.1 63.5% Apr 2025 Official leaderboard
6 Qwen3-VL-235B-A22B 62.9% Sep 2025 Official leaderboard
7 GPT-4o Mar 2025 59.7% Mar 2025 Official leaderboard
8 Qwen2.5-VL-72B 59% Jan 2025 Official leaderboard
9 Gemma 3 27B 57.3% Mar 2025 Official leaderboard
10 Gemma 3 12B 55.3% Mar 2025 Official leaderboard
11 Qwen2.5-VL-7B 50.9% Jan 2025 Official leaderboard
12 Gemma 3 4B 50.7% Mar 2025 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 Text-to-Image, Image Editing and Interleaved. Gemini 3 Pro leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.

Text-to-Image

12 models we track have a score here.

  1. 1 Gemini 3 Pro 74.4%
  2. 2 Gemini 2.5 Pro Jun 2025 ↑1 70.5%
  3. 3 GPT-5 ↓1 70.5%
  4. 4 GPT-4.1 ↑1 65.8%
  5. 5 Gemini 2.5 Flash Sep 2025 ↓1 63.1%
  6. 6 Qwen3-VL-235B-A22B 62%
  7. 7 GPT-4o Mar 2025 60.3%
  8. 8 Qwen2.5-VL-72B 59.1%
  9. 9 Gemma 3 27B 58.3%
  10. 10 Gemma 3 12B 56%

Number one over time

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

Google DeepMind
Own leaderboard Not tested since Dec 2025
Gemini 3 Pro 76.3%

Source: Official Multimodal RewardBench 2 leaderboard

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
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemini 3 Pro GPT-5 GPT-4.1 Qwen2.5-VL-72B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemini 3 Pro GPT-4.1

Source: Official Multimodal RewardBench 2 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.

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.

Meta Google DeepMind
Multimodal RewardBench 2 Not tested since Dec 2025
Gemini 3 Pro 76.3%
Multimodal RewardBench Not tested since Feb 2025
Multimodal RewardBench 2 Not tested since Dec 2025
Gemini 3 Pro 76.3%

Source: Official leaderboards

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

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 12 months ago, in November 2025, for Gemini 3 Pro. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (15 entries).
  • Pass Models scored 12 models we track, plus 3 other entries on its leaderboard.
  • Warning Newest score The newest came 12 months ago, in November 2025, for Gemini 3 Pro.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Pro (#56).
  • Pass Headroom The best is 76.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 12 rows for models we track and 3 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