Multimodal & Grounded benchmark
ERQA leaderboard
Every model the catalog carries a published ERQA value for, ranked by that value.
CategoryMultimodal & Grounded
MeasureGrounded image reasoning
TasksEvidence-based visual QA
DifficultyGrounded multimodal reasoning
Published byQwen3.6 launch benchmarks
A grounded visual reasoning benchmark focused on evidence-based question answering over real images.
ERQA ranking
13 models with a published ERQA value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Grounded image reasoning |
|---|---|---|---|
| 1 |  Qwen3.8 Max | Alibaba | 77.8 |
| 2 | Qwen3.8-Flash-Next | Alibaba | 72.3 |
| 3 |  Seed 2.1 Pro | ByteDance | 72.0 |
| 4 | Seed 2.1 Turbo | ByteDance | 71.3 |
| 5 | Qwen3.8-Omni-Flash | Alibaba | 71.0 |
| 6 | Qwen3.7 Plus | Alibaba | 69.8 |
| 7 |  Gemini 3.1 Pro | 69.4 | |
| 8 | Qwen3.8-27B | Alibaba | 65.5 |
| 9 |  GPT-5.4 | OpenAI | 65.4 |
| 10 |  Muse Spark | Meta | 64.7 |
| 11 | Qwen3.6-27B | Alibaba | 62.5 |
| 12 |  Grok 4.20 | xAI | 54.1 |
| 13 |  Claude Opus 4.6 | Anthropic | 51.6 |
Evidence key: Observed
Rows are ordered by the value Qwen3.6 launch benchmarks published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.