Not runnable todayGenerationVerified 2026-11-05
Qwen2.5-VL 7B Instruct
Qwen/Qwen2.5-VL-7B-Instruct
Qwen2.5-VL 7B Instruct is a text and image language model (Apache-2.0). Not runnable on Mixpeek today.
Something blocks it today, such as the license, remote code or a GPU-only runtime. The reason is below.
No demo for Qwen2.5-VL 7B Instruct yet
A video, image, slideshow or playground goes here once we record one on our own data.
What it is good and bad at
- Best at
- Local structured extraction from scraped pages with no provider API calls (web_scraper llm_provider=vllm).
- Worst at
- Runs nowhere today (vllm is not in the engine dependencies); needs half an L4 and 6 GB per actor when it does; the extractor passes page text only, so the vision side is unused.
- Task
- Generation
- Input
- Text, Image
- Output
- Text, Structured data
- Use cases
- Captioning, OCR, Tagging
Templates that use it
No template uses this model yet. Browse templates.
Input and output
Input
- text
- max_tokens
- 8192 (vllm_max_model_len)
Output
- text
- JSON matching response_shape
How Mixpeek wires it
- Extractors
web_scraper@v1llm_provider=vllm with llm_model=Qwen/Qwen2.5-VL-7B-Instruct (LOCAL_LLM_MODELS)
- Feature URIs
- None today
- Vector indexes
- None
- Retriever stages
- None
- Combines with
- web_scraper crawl -> LLMExtractorProcessor on vLLM -> extracted_data
How it runs
- Status
- Not runnable today web_scraper's vLLM path imports vllm (shared/models/llm/vllm.py), but vllm is not in server/pyproject.toml or poetry.lock, so the import fails and the step cannot run.
- Serve apps
- None
- Runtime
- vllm.LLM (VLLMService, trust_remote_code false, gpu_memory_utilization 0.85)
- Weights
- safetensors, no remote code
- Parameters
- 8,292 M
- Size on disk
- 16,584 MB
- Preprocessing
- Page content and title into the extraction prompt.
- Postprocessing
- Parse JSON against response_shape.
- Serving
- batch, batch 12
- Hardware
- Needs a GPU
Scale to zero
- Minimum replicas
- Not recorded
- Maximum replicas
- Not recorded
- Idle timeout
- Not recorded
- Memory
- 6 GB
- CPUs
- 0.5
- GPUs
- 0.5
- Cold start, node warm
- Not recorded
- Cold start, new node
- Not recorded
Measured cost
Unmeasured
Never run in this codebase's image; nothing measured.
Our evaluations
Unmeasured
We have not evaluated this model on our own data yet. Publisher results are in the sources below.
Questions
- What is Qwen2.5-VL 7B Instruct best at?
- Local structured extraction from scraped pages with no provider API calls (web_scraper llm_provider=vllm).
- What are the limits of Qwen2.5-VL 7B Instruct?
- Runs nowhere today (vllm is not in the engine dependencies); needs half an L4 and 6 GB per actor when it does; the extractor passes page text only, so the vision side is unused.
- Can Qwen2.5-VL 7B Instruct be used commercially?
- Yes. Apache-2.0 weights; no use restrictions.
- How does Qwen2.5-VL 7B Instruct run on Mixpeek?
- web_scraper's vLLM path imports vllm (shared/models/llm/vllm.py), but vllm is not in server/pyproject.toml or poetry.lock, so the import fails and the step cannot run.