RealWorldQA
We show this table for reference; we do not rank on it.
A grounded visual QA benchmark focused on answering practical questions about real-world images and scenes.
Benchmark score on RealWorldQA — November 13, 2026
We compile the RealWorldQA rows from provider self-reports and secondary reports. Qwen3.8-Flash-Next leads the table at 88.5%, followed by Qwen3.8 Max (88.0%) and Qwen3.8-Omni-Flash (87.7%). We do not use these results to rank models overall.
Qwen3.8-Flash-Next
Alibaba
Reported by AlibabaQwen3.8 Max
Alibaba
Reported by AlibabaQwen3.8-Omni-Flash
Alibaba
Reported by Alibaba18 modelsMultimodal & GroundedCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (18 models)
None of the 18 results on this page has a documented run by the benchmark or an independent evaluator: 14 reported by the labs that built the models and 4 reported in another publisher’s document.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Qwen3.8-Flash-NextAlibabaReported by Alibaba | 88.5% | Reported by Alibaba | Not reported | Open |
| 2 | Qwen3.8 MaxAlibabaReported by Alibaba | 88.0% | Reported by Alibaba | Not reported | Open |
| 3 | Qwen3.8-Omni-FlashAlibabaReported by Alibaba | 87.7% | Reported by Alibaba | Not reported | Closed |
| 4 | Qwen3.7 PlusAlibabaReported by Alibaba | 86.9% | Reported by Alibaba | Not reported | Closed |
| 5 | Qwen3.8-27BAlibabaReported by Alibaba | 85.9% | Reported by Alibaba | Not reported | Open |
| 6 | Qwen3.6-35B-A3BAlibabaReported by Alibaba | 85.3% | Reported by Alibaba | Not reported | Open |
| 7 | Qwen3.6-27BAlibabaReported by Alibaba | 84.1% | Reported by Alibaba | Not reported | Open |
| 8 | Qwen3.5 397BAlibabaReported by Alibaba | 83.9% | Reported by Alibaba | Not reported | Open |
| 9 | Gemini 3 ProGoogleAnother lab’s comparison table | 83.3% | Another lab’s comparison table | Not reported | Closed |
| 10 | GPT-5.2OpenAIAnother lab’s comparison table | 83.3% | Another lab’s comparison table | Not reported | Closed |
| 11 | Kimi K2.5Moonshot AIAnother lab’s comparison table | 81.0% | Another lab’s comparison table | Not reported | Open |
| 12 | Ternary Bonsai 2 27BPrism MLReported by Prism ML | 80.1% | Reported by Prism ML | Not reported | Open |
| 13 | LFM3-VL-2BLiquidAIReported by LiquidAI | 79.3% | Reported by LiquidAI | Not reported | Open |
| 14 | Claude Opus 4.5AnthropicAnother lab’s comparison table | 77.0% | Another lab’s comparison table | Not reported | Closed |
| 15 | LFM2.5-VL-3BLiquidAIReported by LiquidAI | 73.1% | Reported by LiquidAI | Not reported | Open |
| 16 | ZAYA1-VL-8BZyphraReported by Zyphra | 65.0% | Reported by Zyphra | Not reported | Open |
| 17 | North Micro Vision InstructCohereReported by Cohere | 62.2% | Reported by Cohere | Not reported | Open |
| 18 | LFM2.5-VL-450MLiquidAIReported by LiquidAI | 58.4% | Reported by LiquidAI | Not reported | Open |
- Lab-reported
- Benchmark-run
- Independent
Among the reported RealWorldQA rows, Qwen3.8-Flash-Next is first at 88.5%. The third row is 0.8 points behind. The broader top-10 range is 5.2 points, so many of the published results sit in a relatively narrow band.
18 models have been evaluated on RealWorldQA. The benchmark falls in the Multimodal & Grounded category. RealWorldQA is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About RealWorldQA
Year
2026
Tasks
Real-world visual question answering
Format
Image-grounded QA
Difficulty
General visual reasoning
RealWorldQA is useful because it emphasizes practical perception and grounded answering on realistic images rather than synthetic or purely academic multimodal tasks.
Freshness and provenance
Version
RealWorldQA 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does RealWorldQA measure?
A grounded visual QA benchmark focused on answering practical questions about real-world images and scenes.
Which model scores highest on RealWorldQA?
Qwen3.8-Flash-Next by Alibaba currently leads with a score of 88.5% on RealWorldQA.
How many models are evaluated on RealWorldQA?
18 AI models have published results on RealWorldQA in the BenchLM catalog.
Compare top models on RealWorldQA
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.