Qwen3.8 Max is a top-tier frontier AI model developed by Alibaba and released in August 2026. It succeeds Qwen 3.7-Max.

Is Qwen3.8 Max available as an open-weight model?

Qwen3.8 Max is a hosted proprietary model, available through Qwen Studio and QwenCloud. Alibaba also has an open-weight sibling, Qwen3.8 2.4T A95B.

What is Qwen3.8 Max's context window?

Qwen3.8 Max supports a context window of about 1M tokens (984K), large enough to hold entire codebases, lengthy technical documents, or extended agentic sessions without truncation. Maximum output is 131,072 tokens.

Is Qwen3.8 Max available in FlowHunt?

Yes — Qwen3.8 Max is directly available in FlowHunt. You can use it to build AI agents, automated workflows, chatbots, and document processing pipelines without any additional setup.

Alibaba Frontier Available in FlowHunt

Qwen3.8 Max

Qwen3.8 Max is a top-tier frontier AI model from Alibaba, released in August 2026. It is a proprietary hosted model, available through Qwen Studio and QwenCloud. The context window is about 1M tokens (984K), with a maximum output of 131,072 tokens.

Alibaba previewed Qwen3.8 Max on July 19, 2026 at the World AI Conference in Shanghai and published the full release, with its own benchmark results, on August 3, 2026. Thinking is always enabled, with low, medium, and xhigh reasoning settings. It accepts text and visual input, and the API follows both OpenAI-compatible and Anthropic-compatible specifications.

Alibaba also has an open-weight sibling, Qwen3.8 2.4T A95B. Qwen3.8 Max itself is served as a hosted model.

Try in FlowHunt Back to AI Leaderboard

Benchmarks

Each score appears once, with its rank among tracked models and who measured it.

Compared with Frontier peers on the same scale

Qwen3.8 Maxthis model
GPT-6.1 Solthis model leads on 0 of 3
Claude Opus 5.5this model leads on 0 of 5

Verified independent evaluator, counted in the ranking Self-reported published by Alibaba, shown but not ranked

Overall

AA Intelligence Index

Independent composite of agent, coding, knowledge and science tests by Artificial Analysis (v4).

Source ↗

Qwen3.8 Max 45.4 GPT-6.1 Sol 51.8 Claude Opus 5.5 57.6

#11of 45 Verified

Coding

DeepSWE

113 long-horizon software tasks written from scratch, graded by hidden tests.

Source ↗Also: Self-reported 56.6% (Alibaba-reported, via secondary write-up; also in GLM-5.3 card)

Qwen3.8 Max 57.5% GPT-6.1 Sol Claude Opus 5.5 74.2%

#12of 17 Verified

SWE-bench Pro

Real issues across several languages and large codebases.

Source ↗

Qwen3.8 Max 67.7% GPT-6.1 Sol Claude Opus 5.5 89.9%

Not ranked Self-reported

SWE-bench Verified

500 verified GitHub issues. Saturated at the frontier.

Source ↗

Qwen3.8 Max 85.6% GPT-6.1 Sol Claude Opus 5.5

#8of 22 Verified

Agents

Terminal-Bench 4.0

Multi-step professional tasks done by an agent in a command-line terminal.

Source ↗

Qwen3.8 Max 38.9% GPT-6.1 Sol 56.1% Claude Opus 5.5 59.6%

#10of 36 Verified

Reasoning

Humanity's Last Exam

Expert-written questions across 100+ subjects, no tools.

Source ↗Also: Self-reported 43.6% (no tools; Alibaba-reported, secondary)

Qwen3.8 Max 43.1% GPT-6.1 Sol 52.9% Claude Opus 5.5 61.4%

#17of 44 Verified

GPQA Diamond

Graduate-level science questions that search cannot answer.

Source ↗Also: Verified 93.7% (Vals AI own implementation)Also: Self-reported 92.6% (Alibaba-reported, secondary)

Qwen3.8 Max 92.8% GPT-6.1 Sol Claude Opus 5.5

#11of 37 Verified

What to Use Qwen3.8 Max For

Terminal & Agentic Automation

Alibaba's own results point to strength on terminal agents and general agent tasks, making it a fit for automation that runs shell and tool workflows.

Software Engineering

It handles coding and repository-level tasks. Alibaba's published results show it is weaker on the hardest software engineering benchmarks, so test it on your own code first.

Multimodal & Long-Document Work

It accepts text and visual input and has a context window of about 1M tokens, which suits document understanding and long-input analysis in a single prompt.

Strengths and limitations

Strengths

Terminal and agentic automation
Multimodal work with text and visual input
Instruction following
Selectable reasoning effort (low, medium, xhigh)
Very long documents and large codebases (about 1M context)

Limitations

Hosted and closed-weight, so it cannot be self-hosted
Thinking is always enabled and cannot be turned off
Weaker on the hardest software engineering benchmarks in Alibaba's own results

Use it in FlowHunt

Use Qwen3.8 Max in any flow, next to your own tools and knowledge sources.

1Open a flow, or create one from a template
2Add an AI Agent or Generator component
3Pick Qwen3.8 Max in the model selector

Try in FlowHunt

Related AI Models

About Alibaba Hangzhou, China · Founded 1999 · qwenlm.github.io

Alibaba Group is one of China's largest and most diverse technology conglomerates, founded in 1999 by Jack Ma and headquartered in Hangzhou. Through its DAMO Academy and Tongyi Lab, Alibaba has become a major force in large language model development with the Qwen (Tongyi Qianwen) family. The Qwen 3.7 generation, released in mid-2026, marked Alibaba's transition from domestic deployments to international benchmark competitiveness, and the Qwen 3.8 generation followed in 2026. Alibaba's AI strategy is defined by open-weight releases as a counterbalance to US API-only models, integration of generative AI across its cloud and e-commerce businesses, and heavy investment in AI infrastructure across its Alibaba Cloud division.

qwenlm.github.io ↗

Browse the Full AI Model Leaderboard

Compare every tracked model on coding, agent and reasoning benchmarks, with each score marked verified or self-reported.

← Back to AI Leaderboard Try in FlowHunt

Frequently asked questions

: Qwen3.8 Max is a top-tier frontier AI model developed by Alibaba and released in August 2026. It succeeds Qwen 3.7-Max.
: Qwen3.8 Max is a hosted proprietary model, available through Qwen Studio and QwenCloud. Alibaba also has an open-weight sibling, Qwen3.8 2.4T A95B.
: Qwen3.8 Max supports a context window of about 1M tokens (984K), large enough to hold entire codebases, lengthy technical documents, or extended agentic sessions without truncation. Maximum output is 131,072 tokens.
: Yes — Qwen3.8 Max is directly available in FlowHunt. You can use it to build AI agents, automated workflows, chatbots, and document processing pipelines without any additional setup.

Qwen3.8 Max