Alibaba Cloud: Qwen3.8-Flash
qwen3.8-flash- Input / Output
- $0.15 / $0.47
- Context
- 992K
- Max output
- 128K
- Modalities
- VisionChat
- Capabilities
- Tool usePrompt Cache
About Qwen3.8-Flash
Qwen3.8-Flash is Alibaba's price-performance model in the Qwen3.8 generation: a multimodal mixture-of-experts system with a 125-billion-parameter core plus 51 billion N-gram embedding parameters, with 6 billion active per token. Alibaba calls it an early preview of the Qwen4 architecture. It reads text and images and targets agentic coding, long-horizon agent tasks, and mobile-interface work.
Where it works well
- Only 6 billion parameters are active per token, which Alibaba pairs with a design aimed at low cost and latency.
- Hybrid attention mixes Gated DeltaNet with Qwen Sparse Attention to keep long inputs workable.
- Alibaba evaluates it on SWE-bench Pro, Toolathlon Verified, AndroidWorld, and MathVision, covering coding, tools, mobile agents, and visual math.
- Image input plus tool use, JSON output, and prompt caching make it a general agent model.
When to choose another model
- Qwen3.8-Max is stronger on the hardest coding and research tasks and is the better pick when quality matters more than speed.
- It understands images but does not generate them, and it takes no audio input.
- As a preview of a new architecture, behavior may shift between releases, so pin evaluations to your own tests.
Getting started
Create API key
Create a key in Console, then use it with every model on the platform.
Send your first request
Copy the example for your language and run it against the endpoint.
Text to textResponses APIPOST/v1/responsesAPI format:curl https://api.tokenlab.sh/v1/responses \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk-xxx" \ -d '{ "model": "qwen3.8-flash", "input": "Hello!" }'
Pricing
The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.
cacheMode · implicit
per 1M tokens- Official price
- Input $0.15 / Output $0.47 / Cache read $0.016 / Text Input $0.15 / Text Input · Cache read $0.016 / Image Input $0.15 / Text Output $0.47
- Verified price
- —
- Discount
- —
cacheMode · explicit
per 1M tokens- Official price
- Input $0.15 / Output $0.47 / Cache read $0.016 / Cache write $0.20 / Text Input $0.15 / Text Input · Cache read $0.016 / Image Input $0.15 / Text Output $0.47
- Verified price
- —
- Discount
- —
| Official priceper 1M tokens | Verified priceper 1M tokens | Discount | |
|---|---|---|---|
| cacheMode · implicit | Input $0.15 / Output $0.47 / Cache read $0.016 / Text Input $0.15 / Text Input · Cache read $0.016 / Image Input $0.15 / Text Output $0.47 | — | — |
| cacheMode · explicit | Input $0.15 / Output $0.47 / Cache read $0.016 / Cache write $0.20 / Text Input $0.15 / Text Input · Cache read $0.016 / Image Input $0.15 / Text Output $0.47 | — | — |
Usage & activity
Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.
Usage & availability
Last 24 hours- Requests
- Success rate
- P95 latency
- Total tokens
- 30-day success rate
- 100.0%
- 30-day requests
- 100+
- Last active
- 4 days ago
Data is based on aggregate user requests, excluding status checks.
Open in Console
Open Qwen3.8-Flash in Console with a prompt ready to edit or send.
Help me try qwen3.8-flash with a short message at /v1/responses. Show the reply, latency, and cost.
Use cases
Best for- Vision
High-volume agent steps
Use it for the many cheap tool-calling turns in an agent loop, escalating only the hard decisions to a larger model.
Mobile and GUI automation
Feed phone or app screenshots and have it choose taps and inputs, in the style of the AndroidWorld evaluation.
Coding assistants with images
Paste a design screenshot or a bug photo with a code snippet and get a fix or a first implementation.
Visual math tutoring
Photograph a worked problem or graph and ask for a step-by-step check of the reasoning.
Prompt examples
This screenshot shows our settings page on Android. Write the steps to turn off notifications, one action per line.
This mockup image comes with our component library. Produce the HTML and CSS that match it.
Check my solution in this photo of a geometry proof and mark the first wrong step.
This model has conditional pricing. Monthly token totals alone cannot produce a reliable estimate; use the detailed pricing for the request specification, cache, and applicable time window.
FAQ
What is Qwen3.8-Flash exactly?
It is Alibaba's efficiency-focused Qwen3.8 model. A mixture-of-experts design with 125 billion core parameters and 6 billion active per token, it supports text and image input and is aimed at agentic coding and long-horizon agent work at lower cost than Max.
Is Qwen3.8-Flash open-weight or hosted?
Alibaba released a preview of this architecture with open weights, announced as Qwen3.8-Flash-Next, and offers the production Qwen3.8-Flash through its cloud API. The two are separate releases with separate model cards.
What does the Qwen4 preview mean for me?
Alibaba describes the model as an early look at the architecture planned for Qwen4. In practice that means newer techniques such as N-gram embeddings and sparse attention, and possible behavior changes in later versions.
Does Qwen3.8-Flash support tool calling?
Yes. Tool use, JSON schema output, and prompt caching are supported, and Alibaba benchmarks it on Toolathlon Verified for tool use. Define your tool schemas clearly and test multi-step sequences before production.
Should I use Flash or Max?
Use Flash for frequent, latency-sensitive turns and routine coding. Use Qwen3.8-Max for the hardest multi-day coding runs and professional documents. Many teams split work by task difficulty and send only failures upward.
How much does Qwen3.8-Flash cost?
On TokenLab, Qwen3.8-Flash costs Input $0.15 / Output $0.47 per 1M tokens. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.
What are the context window and output limit of Qwen3.8-Flash?
Qwen3.8-Flash accepts up to 991,808 tokens of context and returns up to 131,072 tokens in one response.
Which endpoint should Qwen3.8-Flash use?
Use https://api.tokenlab.sh/v1/responses for Qwen3.8-Flash. The request example below shows the matching code shape.
Which operations does Qwen3.8-Flash support?
Qwen3.8-Flash supports Text to text. Select an operation above to see its endpoint and request example.
Compare Qwen3.8-Flash
Sources
Reviewed Oct 2, 2026
