TallyQA
Amazon Nova Premier 61.5% leads since Apr 2025 Dropped Models answer open-ended counting questions about images, including complex ones that need object relationships or attributes. Scored by exact count accuracy.
In the authors’ words
Most counting questions in visual question answering (VQA) datasets are simple and require no more than object detection. Here, we study algorithms for complex counting questions that involve relationships between objects, attribute identification, reasoning, and more. To do this, we created TallyQA, the world's largest dataset for open-ended counting. We propose a new algorithm for counting that uses relation networks with region proposals. Our method lets relation networks be efficiently used with high-resolution imagery. It yields state-of-the-art results compared to baseline and recent systems on both TallyQA and the HowMany-QA benchmark.
- Level
- Everyday Most adults, without training
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Short answer A number, word or short string matched to a key.
Amazon Nova Premier leads with 61.5% among models released by Apr 2025
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in Apr 2025, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 1 Amazon Nova Premier 61.5%
- 2
Gemma 3 27B 54.3% - 3 Gemma 3 12B 51.8%
- 4 Gemma 3 4B 42.5%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Amazon Nova Premier | 61.5% | Apr 2025 | Model card | |
| 2 | Gemma 3 27B | 54.3% | Mar 2025 | Model card | |
| 3 | Gemma 3 12B | 51.8% | Mar 2025 | Model card | |
| 4 | Gemma 3 4B | 42.5% | Mar 2025 | Model card |
2 models have held number one since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
No launch post has listed it since Apr 2025
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 |
|---|---|---|---|
| 30 Apr 2025 | Amazon Nova Premier | 61.5 | TallyQA |
| 12 Mar 2025 | Gemma 3 27B | 54.3 | TallyQA (PT) |
4 system cards report it, from 2 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 |
|---|---|---|---|---|
| 30 Apr 2025 | Amazon Nova Premier technical report | 61.5 | TallyQA, balanced setting, +/-9.1 (Table 1) | Capabilities |
| 25 Mar 2025 | Gemma 3 4B technical report | 42.5 | TallyQA, PT, 4-shot (Table 11) | Capabilities |
| 25 Mar 2025 | Gemma 3 27B technical report | 54.3 | TallyQA, PT, 4-shot (Table 11) | Capabilities |
| 25 Mar 2025 | Gemma 3 12B technical report | 51.8 | TallyQA, PT, 4-shot (Table 11) | Capabilities |
More in Vision
- MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. GPT-5.1 85.4%
- MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. Kimi K3.1 Pro 84.1%
- CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
- MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy.
Qwen 3.8 Max 97.7% - ChartQA In use Models answer human-written and generated questions about chart images that need visual and arithmetic reasoning. Scored by relaxed accuracy with a 5 percent numeric tolerance.
Command A Vision 90.9% - DocVQA In use Models answer 50,000 questions about 12,000 scanned document images by reading their text and layout. Scoring is ANLS, a string similarity to reference answers.
Seed1.5-VL 96.9%