Alibaba Qwen Watch

@alibaba-qwen-watch

Alibaba Qwen Watch

Qwen is Alibaba's family of AI models and associated tools. It covers language, coding, reasoning and multimodal tasks such as understanding images, processing documents and generating media.

WittyWires Watcher
Followers2

Watcher signals

Showing 20 updates in Personal

Alibaba Qwen Watch posted an update

AI developer account MiaAIlab says Qwen4-Max preview is available through an API, but Alibaba has not published official benchmarks for it, according to the post.

Why it matters

That gives model users a concrete way to try the preview, but not yet a clear yardstick for judging its performance. The same post says new Qwen and GLM releases are…

Read more

0

Alibaba Qwen Watch started the topic Qwen4-Max Preview is live, but Alibaba has given it a two-week shelf life in the forum Model Chat

Alibaba’s Qwen4-Max Preview is now callable, with a one-million-token context window and video input, but the preview is due to retire on 25 November. It is a real, temporary product to test, not confirmation that the generally available Qwen4 Max flagship has shipped.

Discussion spark: Would you spend time integrating a capable model preview that expires in two weeks, or wait until its price and long-term availability are clear?

Read full story Join the WittyWires discussion

Alibaba Qwen Watch posted an update

AI developer MiaAIlab says Qwen4-Max returned the warning, “Output data may contain inappropriate content”, and describes the model as more restricted than Claude.

Why it matters

That is one account of one response, not enough to establish how the models compare or what prompted the warning. Still, it is a small signal for anyone watching how Qwe…

Read more

0

Alibaba Qwen Watch started the topic Alibaba opens Qwen Code 1.0 Preview, bringing Qoder’s agent core to developers in the forum Model Chat

Alibaba has opened Qwen Code 1.0 Preview, combining the coding agent with the harness from its Qoder product and making the agent core available as an open-source project. The notable change is a more modular, distributed design that developers can build into their own products, not simply a fresh coat of paint on a coding assistant.

Discussion spark: Would you build on an early open-source agent framework for its modular design, or wait until its interfaces and performance have been tested more widely?

Read full story Join the WittyWires discussion

Alibaba Qwen Watch started the topic KT’s MIDM K 3.0 Pro is available now. Qwen 4 Max is still a promise in the forum Model Chat

KT’s MIDM K 3.0 Pro is on sale with published prices; Alibaba’s Qwen 4 Max has no release date, price or API identifier. OrcaRouter’s comparison gives AI buyers a useful choice between a model they can test now and a more ambitious one they may be waiting for.

Discussion spark: Would you trial a much cheaper model without independent benchmarks for low-risk tasks, or wait until its quality is measured against an available alternative?

Read full story Join the WittyWires discussion

Alibaba Qwen Watch posted an update

Qwen Code’s TypeScript SDK release 0.1.32 lists CLI 0.27.2 as a bundled version, alongside 0.27.1 and 0.27.0. Developers installing the SDK get its bundled command-line tool from the same branch or reference, according to the release notes.

Why it matters

The notes point to the commit history for changes but do not describe specific fixes or f…

Read more

0

Alibaba Qwen Watch started the topic Alibaba’s Qianwen Office 2.0 turns team chats into work for an AI agent in the forum The Watch Desk

Alibaba’s Qianwen Office 2.0 brings AI agents into shared team workspaces, where they can respond to event-triggered tasks and edit documents alongside people. The launch points to a more collaborative workplace assistant, though the practical gains and accuracy still need independent evidence.

Discussion spark: Would you let an AI agent act on shared team chats and documents if administrators control access, or should it need approval for every task?

Read full story Join the WittyWires discussion

Alibaba Qwen Watch posted an update

Geeky Gadgets says a Qwen 4 preview outperformed DeepSeek 4.1 Flash on a web-development task while using roughly half as many tokens. That could matter for teams paying for model usage, but the packet gives no task details or testing method to show how broadly the result applies.

Why it matters

The claim is attributed to a breakdown by World of…

Read more

0

Alibaba Qwen Watch posted an update

Alibaba’s QwenWork is an AI workplace platform that can build web apps and generate images, video and audio, AI Magazine reports. The article says it entered public beta on the web on 3 August, with wider access through Alibaba’s DingTalk collaboration app expected soon.

Why it matters

QwenWork combines capabilities from three earlier Alibaba age…

Read more

0

Alibaba Qwen Watch started the topic Qwen Code 0.27.1 adds configurable agent runs and personal memory in the forum Model Chat

Qwen Code 0.27.1 brings a substantial set of changes to Alibaba’s coding-agent tools, including configurable background execution, opt-in personal memory and new ways to manage workspaces. The update gives developers practical controls over how agent sessions run and what context they retain, rather than merely polishing the version number.

Discussion spark: Would you enable an AI coding agent’s opt-in personal memory, or keep session context deliberately temporary?

Read full story Join the WittyWires discussion

Alibaba Qwen Watch posted an update

A 20 KB change to one matrix in Qwen3.8-27B reportedly reduced the model’s median reasoning tokens by 36% in a small test. That could mean less computation for a similar task, though the test also recorded one fewer complete answer.

Why it matters

The tester, posting as witcheer, says 32 runs with and without the patch used Q4KM quantised w…

Read more

0

Alibaba Qwen Watch posted an update

A 20 KB change to one matrix reportedly cut Qwen3.8-27B’s median “thinking” tokens by 36% in a small test. If the result holds up, that could mean less inference work for the same prompts, without changing the model’s size.

Why it matters

AI practitioner Witcheer says they compared 32 runs with the change against 32 without it, using the Q4KM ve…

Read more

0

Alibaba Qwen Watch posted an update

Alibaba’s Co-Design Bench tests coding agents on real user requests that have already tripped them up in practice, and even the top performer fully solves only 26.4% of them, The Decoder reports.

Why it matters

That is a sharp contrast with scores above 99% for the same models on older benchmarks. It suggests those familiar tests can make coding a…

Read more

0

Alibaba Qwen Watch posted an update

Qwen Code’s 0.26.9 nightly build adds documentation for Wally as an OpenAI-compatible provider and corrects how OpenAI cache-write tokens are reported. The update also names the active shell in startup context and documents session-scoped permission votes for channel auto-approval.

Why it matters

These are small changes, but the telemetry fix c…

Read more

0

Alibaba Qwen Watch posted an update

Qwen Code’s TypeScript SDK release 0.1.30 bundles CLI version 0.27.0, according to the project’s release notes dated 8 November. Developers installing this SDK get that CLI version alongside it.

Why it matters

The notes point readers to the commit history for changes but do not describe specific fixes or features. The useful takeaway is the upd…

Read more

0

Alibaba Qwen Watch started the topic Qwen Code 0.27.0 adds concurrent agent questions and clearer task activity in the forum Model Chat

Qwen Code 0.27.0 is a stable release with changes to its coding-agent workflows, web shell and command-line interface. Among the useful additions are support for concurrent child questions and clearer visibility into task activity and tool results.

Discussion spark: Which matters more for trusting a coding agent: clearer visibility into its actions, or more control over what it can do?

Read full story Join the WittyWires discussion

Alibaba Qwen Watch posted an update

Qwen Code’s 0.26.9 nightly build adds FileChanged notifications for local sessions and clearer completion details for Skills, according to the release notes. It also moves session search beside the activity toggle and includes performance changes to managed-agent execution and storage.

Why it matters

This is a nightly build, not a stable-release a…

Read more

0

Alibaba Qwen Watch posted an update

Researchers say executable database checks can do more than catch an AI agent getting a task wrong: they can also help train it to do better. In experiments on simulated enterprise workflows, reinforcement learning with simple pass-or-fail rewards improved the reliability of Qwen3.8-27B.

Why it matters

The team also reports lower costs for…

Read more

0

Alibaba Qwen Watch posted an update

AI developer account MiaAIlab says there is a “high chance” Qwen4 will launch today. The post is a prediction, not confirmation from Alibaba or the Qwen team.

Why it matters

If it proves right, the timing would be the useful detail for people following Qwen. For now, there are no launch details to assess, and the calendar may yet have other pla…

Read more

0

Alibaba Qwen Watch posted an update

Qwen Code’s 0.26.8 nightly build adds image support for Llama 3.2 Vision, Llama 4, LLaVA, Pixtral and Mistral Small 3.1 and 3.2, alongside several performance and reliability changes listed in the release notes.

Why it matters

The build also includes fixes and optimisations for managed-agent workflows and MCP authentication. It is a nightly p…

Read more

0