Qwen3 VL 8B Thinking
Qwen · released Oct 11, 2025 · Qwen/Qwen3-VL-8B-Thinking
- Type
- Open weightsApache License 2.0
- Params
- 8.8B
- Context
- 131K
about 98K words of context
Our take
Written Aug 3, 2026Qwen3 VL is a compact vision-language model from Alibaba that can reason through its answers before responding. Its weights are available under a permissive Apache licence, and it can handle long documents of up to 131,072 tokens in a single request.
Pick this for low-cost image-and-text reasoning where you need a permissive licence for local use or fine-tuning. Use it for long-document image analysis at a low parameter count, or when you want to self-host rather than rely on a single provider. Skip it if you need verified quality scores, deep reasoning on complex problems, or predictable throughput from every host.
The case for it
- Cheapest tracked offer is well below the next tier for multimodal inference with reasoning.
- Apache 2.0 licence allows commercial use, fine-tuning and redistribution.
- 131,072-token request limit is large for its parameter class.
The case against it
- No benchmark scores in our data — chat, reasoning, vision and coding performance are all unverified.
- 8.8 billion total parameters with no disclosed mixture-of-experts architecture; reasoning depth may be limited.
- Throughput varies even from the same provider, with no clarity on which speed to expect.
How good is it?
We hold no score for this model.
We look for every model we track on every board we watch, and none of them has turned up Qwen3 VL 8B Thinking — so there is no intelligence, coding, agentic or writing score to show you, not a low one, none.
Can you run it yourself?
- Fits in memory
- weights load entirely on the card
- Spills to system RAM
- some weights offload; much slower
- Too large
- will not load even with offload
- est
- size is calculated; the verdict could change by 10%
Comfortable fit
GeForce RTX 4090 · 24 GB
Room to spare. 15.6 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 23.6 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU) · 16 GB
Room to spare. 4.8 GB spare means a 10% error in the size would not change the answer.
Memory use by level
Against a 24 GB card.
All 3 sizes here are calculated, not measured. We hold no measured file for this model, so each size comes from the parameter count and every verdict above inherits that uncertainty.
Check against your own machine → · All 71 devices, with every size →
Or rent it from someone else
Cheapest of 3 live listings. Picked at the widest standard context we hold, within one quantisation slice, so the numbers beside it are a price one host actually charges.
- per 1M tokens
- $0.12 in / $1.36 out
- Context served
- 131K
- Throughput
- ~132 tok/s
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| Alibaba Cloud | $0.12 / $1.36 | 131K | 132 tok/s | No | Yesunknown period | Unknown |
| OpenRouter | $0.18 / $2.10 | 131K | not measured | Unknown | Unknown | Unknown |
| Alibaba Cloudfp8 | $0.18 / $2.10 | 131K | 138 tok/s | No | Yesunknown period | Unknown |
OpenRouterAcross the 3 listings we hold: 2 say they do not train on prompts, 0 say they do and 1 do not say. 0 appear in the zero-retention registry we check; the rest are unknown to us rather than confirmed either way.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
- ✓
- Supported
- ✗
- Not supported
- Not published
- host gave no parameter list
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| Alibaba Cloud | ✓ | ✓ | ✓ |
| OpenRouter | ✓ | ✓ | ✓ |
| Alibaba Cloudfp8 | ✓ | ✓ | ✓ |
Tool calling: 3 of 3 listings say yes. JSON output: 3 of 3 listings say yes. Strict schema: 3 of 3 listings say yes.
When we formed this view
Dates behind this page
Prices last checked 5d ago
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- No board we watch has turned up a score, so we hold no quality figures at all.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 3 listings do not say whether they train on prompts.
- We hold no cached-input rate for any of its listings.
Licence and identifiers
What the licence allowsApache License 2.0, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Apache License 2.0
Fully permissive: commercial use, redistribution, and derivatives allowed. Requires attribution and a copy of the license. Includes an express patent grant.
Identifiers
- Hugging Face
- Qwen/Qwen3-VL-8B-Thinking
- Architecture
- Dense
- Modality record
- text+image->text
- Catalogue slug
- qwen-qwen3-vl-8b-thinking