Alibaba Cloud (Qwen) · AI model
Qwen3.8-Flash
Lower-cost Qwen3.8 multimodal model with text, image and video input, a 1M-token context window, up to 131,072 output tokens and hybrid thinking; compatible with OpenAI and Anthropic API protocols.
Last verified Available 7 sources
Key specs
- Provider
- Alibaba Cloud (Qwen)
- Family
- Qwen3.8
- API identifier
- qwen3.8-flash
- Availability
- Available
- Release date
- 2026-08-26
- Context window
- 1,000,000 tokens
- Max output
- 131,072 tokens
- API access
- Yes
- Consumer app
- Unknown
Capabilities
Inputs
- Text input Yes
- Image input Yes
- Audio input No Input modalities listed as image, text, video.
- Video input Yes Video up to 2 hours / 2 GB; Alibaba Cloud directs users to qwen3.8-omni-flash to understand sound in a video.
- File input Partial PDF only, via PDF understanding (Chat Completions API, not Responses API); max 150 MB / 256 pages; a per-page document parsing fee applies (Singapore: $0.0033 per page; not listed as a price row).
Outputs
- Text output Yes
- Image output No
- Audio output No
Features
- Reasoning Optional Hybrid thinking mode, enabled by default; can be disabled per request.
- Tool calling Yes
- Web search Yes Built-in web search; Singapore (International) scope.
"Unknown" means we have not verified this capability yet; it does not mean the model lacks it.
Pricing
Estimate your monthly cost →| Dimension | Tier | Price | Since | Verified | Source |
|---|---|---|---|---|---|
| Input tokens | base | $0.15 per 1M tokens | 2026-10-12 | 2026-10-25 | qwen3.8-flash Model Info - Alibaba Cloud Model Studio Singapore (International) list price; no input-length tiers (0 < tokens <= 1M). |
| Cached input tokens | base | $0.016 per 1M tokens | 2026-10-12 | 2026-10-25 | qwen3.8-flash Model Info - Alibaba Cloud Model Studio Implicit cache hit. |
| Cache writes | base | $0.20 per 1M tokens | 2026-10-12 | 2026-10-25 | qwen3.8-flash Model Info - Alibaba Cloud Model Studio Explicit cache creation. |
| Output tokens | base | $0.47 per 1M tokens | 2026-10-12 | 2026-10-25 | qwen3.8-flash Model Info - Alibaba Cloud Model Studio Single output price listed (no separate thinking / non-thinking price). |
| Tool and search requests | base | $10.00 per 1,000 requests | 2026-10-12 | 2026-10-25 | Web search with large models - Alibaba Cloud Model Studio Singapore: 'agent' search strategy, or the web_search tool via the Responses API or Anthropic-compatible method; per 1,000 calls, plus retrieved content billed as input tokens. No fee is listed for the default turbo/max strategies |
| Other | base | $0.016 per 1M units | 2026-10-12 | 2026-10-25 | qwen3.8-flash Model Info - Alibaba Cloud Model Studio Explicit cache read (explicit cache hit). |
Prices are shown in the currency the provider publishes; we never convert currencies.
Benchmarks
| Benchmark | Setting | Score | Reported by | Measured | Verified | Source |
|---|---|---|---|---|---|---|
| CharXiv Reasoning accuracy %, higher is better | reasoning questions (RQ), with code interpreter (CI); effort not stated | 90.6 % | Self-reported | — | 2026-10-25 | Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery. (Qwen blog, September 2026) |
| reasoning questions (RQ), without code interpreter (CI); effort not stated | 84.6 % | Self-reported | — | 2026-10-25 | Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery. (Qwen blog, September 2026) | |
| DeepSWE v1.1 % resolved (pass@1), higher is better | v1.1; effort not stated; best of Claude Code and mini-SWE-agent harnesses, temp 1.0, top_p 0.95, 256K context | 58.7 % | Self-reported | — | 2026-10-25 | Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery. (Qwen blog, September 2026) |
| Humanity's Last Exam accuracy %, higher is better | tools not stated; effort not stated; GPT-4o judge | 35.9 % | Self-reported | — | 2026-10-25 | Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery. (Qwen blog, September 2026) |
There is no maintained public leaderboard (the project site says it is under construction), so almost all current numbers are self-reported by providers. Results with tools such as a code interpreter are much higher than without: Alibaba reports 83.7% without and 90.2% with tools for Qwen3.8-27B. Top models now exceed the human baseline, so differences at the top are small and may reflect label noise. Descriptive-question scores are a different metric and must not be recorded here.
The official leaderboard uses mini-swe-agent for all models and shows whole-number percentages with margins of about ±1 to ±6 points. Providers often report scores from their own scaffolds or effort settings; DeepSeek reports 65.5% to 74.2% for the same model depending on the scaffold. DeepSWE v1 scores are not comparable with v1.1. With only 113 tasks, small differences are within noise.
Results differ a lot by setting: with tools versus without tools (Anthropic reports 45.9% without tools and 57.4% with tools for Claude Haiku 5.5), and the full set versus the text-only subset. Providers use different judge models and harnesses (Scale uses o3-mini; Alibaba states GPT-4o). Derived sets such as HLE-Verified (a third-party revision with 1,811 scored items), HLE-Rolling and HLE-Diamond are different item sets and are not comparable. The questions are public, so contamination is possible, and community reviews have reported flawed items.
One row per result, without rank numbers: each result was measured under its own setting (effort level, harness, tools, test version), as stated by the source, and scores compare only within the same setting. Self-reported results come from the provider and have not been reproduced independently.
Rankings
- #15 of 26 in Best AI for students as of 2026-10-26
Derived from our published ranking snapshots; each ranking explains its criteria and weights.
Compare Qwen3.8-Flash
Guides
- AI token pricing explained: input, output, cached and batch rates
- Reasoning models explained: thinking modes, effort and cost
- Multimodal AI explained: image, audio, video and file inputs, and what they cost
- Context windows explained: limits, long-context pricing and recall
- How to choose an AI model: a practical decision framework
About Qwen3.8-Flash
qwen3.8-flash is the fast, lower-priced model of the Qwen3.8 series on Model Studio. It accepts text, images and video and supports function calling, structured output, web search and context caching. Thinking mode is enabled by default and can be disabled per request. The official Hugging Face card for Qwen3.8-Flash-Next states that Qwen3.8-Flash is the hosted version based on that open-weights model, adding production features such as a 1M default context and built-in tools.
Strengths and weaknesses
Strengths
- 1,000,000-token context window with up to 131,072 output tokens and up to 262,144 chain-of-thought tokens
- Text, image and video input; PDF understanding supported in Singapore
- List price of USD 0.15 input / USD 0.47 output per 1M tokens, with implicit and explicit context caching
- Compatible with OpenAI and Anthropic API protocols
Weaknesses
- Batch inference is not supported in the Singapore region
- No audio input (audio understanding is offered by qwen3.8-omni-flash)
Sources
- qwen3.8-flash Model Info - Alibaba Cloud Model Studio
- Model lifecycle and updates - Alibaba Cloud Model Studio
- Visual understanding - Alibaba Cloud Model Studio
- PDF file understanding via VL model - Alibaba Cloud Model Studio
- Use deep thinking models via API | Model Studio - Alibaba Cloud Model Studio
- Web search with large models - Alibaba Cloud Model Studio
- Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery. (Qwen blog, September 2026)
Last verified Page updated 2026-10-26