Gemini Flash vs Qwen-VL
Pick Gemini Flash for cheap, high-volume PDF to Markdown or JSON extraction through a hosted API. Pick Qwen-VL if you need open weights under Apache-2.0, self-hosting, or layout output with bounding boxes. Qwen-VL can loop and repeat text on low-quality scans, so add loop checks before using it for faithful transcription.
Gemini Flashgoogle/gemini-flash | Qwen-VLqwen/qwen-vl | |
|---|---|---|
| Type | Vision model | Open source |
| Price | 15 credits / page | 2 credits / page |
| Per 1,000 pages | $15.00 | $2.00 |
| Typical speed | 1.6 s / page | 2.0 s / page |
| Best for | Handwriting and mixed layouts | Flexible extraction with prompts |
| Inputs | PDF, images | PDF, images |
| Languages | 100+ languages | 30+ languages |
| Max pages per request | 1,000 | 200 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | ✓ Yes |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | No | ✓ Yes |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | ✓ Yes | ✓ Yes |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose Gemini Flash if
- Gemini 3 Flash scores 92.62 on OmniDocBench v1.6, close to Gemini 3 Pro at 92.91.
- A context window of up to 1M tokens fits long documents in one request.
- It is billed at $0.50 per 1M input tokens and $3.00 per 1M output tokens, or $0.25 and $1.50 with Batch/Flex, which suits bulk PDF to Markdown.
Choose Qwen-VL if
- Weights are Apache-2.0 and come in sizes from 2B to 235B-A22B MoE, so you can self-host on your own hardware.
- It outputs bounding boxes and a Qwen HTML document format, not just plain text.
- It does OCR in 32 languages, up from 10 in the previous generation, and handles photos and screenshots with text.
Key differences
| Gemini Flash | Qwen-VL | |
|---|---|---|
| Licence | Proprietary | Apache-2.0 (Qwen3-VL repository) |
| How you run it | Hosted API only, through the Gemini API or Vertex AI. Not for air-gapped or on-prem use. | Self-hosted open weights, or hosted on Alibaba Cloud. |
| Price | $0.50 per 1M input tokens and $3.00 per 1M output tokens (Gemini 3 Flash Preview). Cost per page depends on page density and output length. | Free to self-host, you pay for compute. Hosted pricing on Alibaba Cloud: Not published. |
| Output formats | Markdown, JSON, HTML, plain text. Output is prompt-driven, with no fixed schema or confidence scores. | Plain text, HTML, bounding boxes, JSON, Markdown. |
| Benchmark | 92.62 on OmniDocBench v1.6 (Gemini 3 Flash). | 64.6 on olmOCR-bench (Qwen 3 VL 8B), below dedicated OCR models. Not published for OmniDocBench. |
| Hardware | None to manage, since it is hosted. | FP8 checkpoints need an H100-class GPU and CUDA 12+. The 235B example uses 8-way tensor parallel on vLLM. |
| Known failure | Newer Flash models move from $0.75/$3.75 to $1.50/$7.50 per 1M tokens on January 1, 2027. | Can loop and repeat the same text, worse with greedy decoding and on low-quality scans. One report saw loops on 10-37% of OCR outputs versus 2-3% baseline after a transformers 5.17 update. |
Sources
- Gemini API pricing
- OmniDocBench leaderboard
- Gemini pricing changes 2026
- Qwen3-VL GitHub
- Issue #1611 infinite repetition
- olmo-eval issue #441
What each one is
Gemini Flash
General vision model with strong handwriting and layout understanding at a low price.
Gemini Flash detailsQwen-VL
Alibaba's open-weight vision language models (Qwen2.5-VL, Qwen3-VL). Strong general OCR and grounding.
Qwen-VL detailsTry both on your own file
Run the file once with each model and compare the output.
PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?
Use one, fall back to the other
{"model": "google/gemini-flash","fallbacks": ["qwen/qwen-vl"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Gemini Flash vs Granite DoclingGemini Flash vs Hunyuan OCRGPT vs Qwen-VLMistral OCR vs Qwen-VLQianfan OCR vs Qwen-VLGemini Flash vs dots.ocrGemini Flash vs LlamaParseOCRmyPDF vs Qwen-VL
Gemini Flash alternativesQwen-VL alternativesBest OCR models in 2026