Is Qwen 3.8 Flash available with Zero Data Retention on Elosia?

Qwen 3.8 Flash runs on Elosia's standard backend; ZDR is available on many other models in the catalog.

How much does Qwen 3.8 Flash cost on Elosia?

Qwen 3.8 Flash is billed by usage at $0.15 per 1M input tokens and $0.47 per 1M output tokens, within your Elosia plan.

What is Qwen 3.8 Flash's context window?

Qwen 3.8 Flash supports a context window of up to 1,000,000 tokens.

Is Qwen 3.8 Flash hosted in the EU?

Yes, Qwen 3.8 Flash is available on Elosia's EU-hosted (EdenAI) backend for enterprise organizations.

When should you choose Qwen 3.8 Flash?

Qwen 3.8 Flash is a balanced model: a strong quality/cost trade-off for everyday production work.

Qwen 3.8 Flash

QwenBalanced

ThinkingTool UseVisionStructured Output

About this model

Qwen3.8 Flash is a multimodal reasoning model from Alibaba. It is suited for coding assistance, agentic workflows, visual understanding, document and codebase analysis, desktop interaction, chart analysis, and long-video analysis.

Performance Tier

Balanced

Qwen 3.8 Flash is a balanced model from Qwen : strong performance at a reasonable price.

Strong cost-performance ratio. Reliable for most professional use cases without premium pricing.

Pricing

This model is included in Elosia plans

Eco

Minimal cost. Ideal for very high volume or simple tasks.

Type	per 1M tokens
Input (prompt)	$0.150
Output (completion)	$0.470
Cache read	$0.016

Capabilities

Context Length1.0M

Max Output Tokens131K

TokenizerQwen

Inputtext, image, video

Outputtext

Release DateAugust 26, 2026

Benchmarks

General Intelligence

MMLU

Not reported

GPQA Diamond

91.7%

Mathematics

MATH-500

Not reported

Programming

HumanEval

Not reported

SWE-bench Multilingual

81%

LiveCodeBench

91.9%

Reasoning

Humanity's Last Exam

35.9%

Multimodal

MMMU-Pro

79.8%

Agentic

SWE-bench Pro

62.5%

Terminal-Bench 2.1

86.1%

Where does Qwen 3.8 Flash stand?

Compare its performance index against every other model.

View the performance leaderboard

Recommended Use Cases

CodingAnalysisData ExtractionSummarizationGeneral Chat

Strengths

Sparse mixture of experts with 125B parameters and only 6B active per token, drawn from 512 experts, plus a separate n-gram embedding table that scales capacity without adding compute
Hybrid attention pairing Gated DeltaNet with Qwen Sparse Attention, which selects micro-blocks rather than single tokens to cut long-context latency
Native image and video input through a built-in vision encoder, trained for GUI agents on desktop, mobile and web as well as chart and document analysis
Built for agents: tool calling, switchable thinking, and reasoning preserved across turns by default for consistent multi-step decisions
Open weights in BF16 and FP8, served by vLLM, SGLang and TokenSpeed

Limitations

Limited factual knowledge density, measured independently, in line with its small active parameter count
The Qwen Community License requires a separate agreement for commercial model-as-a-service use or for AI coding and office assistants
Published scores cover the open-weight Flash-Next checkpoint, an experimental preview of the architecture; the API version adds production features without its own benchmark table
Qwen publishes no Terminal-Bench 2.1 or MMMU-Pro score; the ones shown come from third-party testing