Alibaba Unveils Qwen3.8-max: a Cost-effective AI Powerhouse

WHO:

Techmeme - Alibaba Group is the organization releasing the Qwen3.8-Max model.

MarkTechPost - The Qwen team at Alibaba is responsible for the development of Qwen3.8-Max.

WHAT:

Techmeme - Alibaba's Qwen team moved Qwen3.8-Max from preview to general availability.

MarkTechPost - Qwen3.8-Max is described as the most capable model in the Qwen family to date.

Techmeme - Qwen3.8-Max is priced at $2 per 1 million input tokens and $6 per 1 million output tokens, undercutting Kimi K3's pricing of $3 per 1 million input tokens and $15 per 1 million output tokens.

Techmeme - Alibaba plans to release the weights of Qwen3.8-Max and Qwen3.8-27B next week.

VentureBeat - Qwen3.8-Max scores 86.1 on the OSWorld-Verified benchmark, outperforming GPT-5.6 Sol Max (83.2) and Fable 5 (85.0).

VentureBeat - Qwen3.8-Max also leads on PaperBench with a score of 93.0, TerminalBench 2.1 with 86.6, Vision2Web with 69.0, and LVBench with 81.8.

VentureBeat - Qwen3.8-Max is positioned as capable of autonomously completing software projects lasting more than 10 days and performing iterative chip-design optimization.

Techmeme - Alibaba made Qwen3.8-Max widely available on Monday through application programming interfaces.

The Verge - Alibaba claims Qwen3.8-Max's performance rivals the best systems from US frontier labs Anthropic and OpenAI, as well as domestic rivals like Moonshot AI’s Kimi K3.

The Verge - Results from Alibaba’s own testing suggest Qwen3.8-Max's performance broadly matches — and sometimes exceeds — that of Fable 5 on benchmark tests.

The Verge - On the Arena text model leaderboard, Qwen3.8-Max trails only Fable 5 and three models in Anthropic’s Opus family.

The Verge - For frontend coding, Qwen3.8-Max is beaten only by two Claude Opus models and Kimi K3, and for visual analysis, only Fable 5 trounces it.

The Verge - Qwen3.8-Max marks a return to open-weight releases for Alibaba after a brief pivot towards proprietary releases for its more advanced models earlier this year.

The Verge - Open-weight releases have become a growing point of differentiation for China’s AI industry, where they have become the norm.

The Verge - The release of Qwen3.8-Max closely follows the release of Kimi K3, viewed as another challenge to American AI dominance.

Techmeme - Alibaba has priced Qwen3.8-Max at $2 per 1 million input tokens and $6 per 1 million output tokens via its API.

MarkTechPost - Qwen3.8-Max is a 2.4 trillion parameter MoE model and is described as the most capable model in the Qwen family to date.

Techmeme - Qwen3.8-Max is said to outperform Kimi K3 on some benchmarks.

Techmeme - Alibaba plans to release the weights of Qwen3.8-Max and Qwen3.8-27B next week.

WHERE:

Techmeme - The Qwen3.8-Max model is made available through application programming interfaces.

WHEN:

MarkTechPost - The open weights for Qwen3.8-Max are expected to be published next week.

WHY:

Techmeme - The pricing strategy for Qwen3.8-Max is positioned below that of Kimi K3 to enhance competitiveness in the market.

HOW:

VentureBeat - Qwen3.8-Max is described as an autonomous coworker capable of executing projects that span days rather than minutes.

MarkTechPost - Qwen3.8-Max accepts text, image, and video input across a 1 million token context.

(UPDATED August 4, 2026 10:34 UTC)

Alibaba's Qwen team has officially moved Qwen3.8-Max from preview to general availability, marking it as the most capable model in the Qwen family to date. The pricing for Qwen3.8-Max is set at $2 per one million input tokens and $6 per one million output tokens, which undercuts the pricing of Kimi K3. Additionally, Alibaba plans to release the weights of both Qwen3.8-Max and Qwen3.8-27B next week.

(UPDATED August 4, 2026 04:08 UTC)

Qwen3.8-Max has made headlines by scoring an impressive 86.1 on the OSWorld-Verified benchmark, surpassing both GPT-5.6 Sol Max and Fable 5. It also leads in several other performance metrics, including a score of 93.0 on PaperBench and 86.6 on TerminalBench 2.1. This advanced model is positioned as an autonomous coworker, capable of completing software projects that last more than ten days and optimizing chip designs iteratively.

(UPDATED August 3, 2026 16:23 UTC)

Alibaba has made its Qwen3.8-Max widely available through application programming interfaces, marking a return to open-weight releases after a brief shift towards proprietary models earlier this year. The company claims that Qwen3.8-Max's performance rivals top systems from US frontier labs like Anthropic and OpenAI, and it has shown competitive results against domestic rivals such as Moonshot AI's Kimi K3. Additionally, Qwen3.8-Max ranks just behind Fable 5 on the Arena text model leaderboard and is only surpassed in frontend coding by two Claude Opus models and Kimi K3.

(UPDATED August 3, 2026 12:56 UTC)

Alibaba Group has announced the release of its new Qwen3.8-Max model, developed by the Qwen team, which boasts an impressive 2.4 trillion parameters and is considered the most capable model in the Qwen family to date. Priced at $2 per 1 million input tokens and $6 per 1 million output tokens, Qwen3.8-Max will be available through application programming interfaces and is expected to outperform Kimi K3 on some benchmarks. Additionally, Alibaba plans to publish the open weights for Qwen3.8-Max next week, enhancing its competitiveness in the market.