Alibaba has announced Qwen3.8-Max as its new flagship model, describing it as a 2.4T-parameter sparse architecture focused on long-horizon agentic work, coding, and multimodal reasoning. The company confirmed that open weights for Qwen3.8-Max will be released next week alongside the open-weight Qwen3.8-27B variant.
Alibaba's Qwen team has announced Qwen 3.8 Max, a new 2.4T-parameter flagship model focused on coding, long-horizon agentic work, and multimodal reasoning. The company confirmed that open-weight versions of both Qwen 3.8 Max and the smaller Qwen 3.8-27B will be released next week.
Alibaba's Qwen team has made Qwen3.8-Max broadly available via API, with open weights for both Qwen3.8-Max and the smaller Qwen3.8-27B checkpoint scheduled to ship next week. The flagship model is a 2.4-trillion-parameter mixture-of-experts architecture that accepts text, image, and video inputs.
Google has released three new models: Gemini 3.6 Flash for efficient general-purpose agent workloads, 3.5 Flash-Lite for low-latency applications, and a cyber-specialized 3.5 Flash model integrated with CodeMender.
On July 19, Alibaba's Qwen team previewed Qwen3.8-Max-Preview, a new flagship multimodal model with 2.4 trillion parameters. The announcement was made during the World AI Conference in Shanghai, arriving two days after Moonshot AI released its Kimi K3 model.
Artificial Analysis has updated its agentic index to rank Qwen 3.8 Max as the best overall model, placing it ahead of Opus 5.
This newsletter covers recent developments in the AI industry, including significant pricing shifts by OpenAI, new model releases from Alibaba, and major leadership changes at Google DeepMind.
The Qwen3.8-2.4T-A95B model, also known as Qwen3.8-Max, is scheduled for an open release next Wednesday.
This AI news roundup for August 3-4, 2026, covers significant model releases from Alibaba Qwen, NVIDIA, Mistral AI, and Black Forest Labs, alongside developments in agent routing, infrastructure, and cybersecurity.
Qwen 3.8 Max achieves a score of 1588 on the Debate Benchmark, improving upon Qwen 3.7 Max's score of 1462. This benchmark evaluates how well large language models hold an argument under adversarial, multi-turn opposition across various topics.
Qwen introduces Qwen-CUA, a native computer-use agent built on a 397B-A17B mixture-of-experts backbone that operates via screenshots and input events without relying on DOM trees or accessibility metadata. The system utilizes a scaffold to maintain up to 20 active screenshots and was trained using a cloud rollout fleet with nearly 100,000 vCPUs across approximately 40,000 verifiable tasks.
Researchers have introduced Qwen-CUA, a native computer-use agent built on a 397B-A17B Qwen mixture-of-experts backbone. The system operates by observing screenshots and acting through keyboard and mouse events, without relying on DOM trees or accessibility metadata.
The article highlights Qwen3.8-Max's performance in a one-shot evaluation context, noting its ability to handle complex physical simulations.
Qwen3.8-Max (2.4T) is a new open-weight model that performs closely to Kimi K3 and DeepSeek V4 Flash across all benchmark categories, while demonstrating superior performance in coding and software tasks.
Alibaba's Qwen team has announced that the next iteration of its model, version 3.8, will be released with open weights in the coming week.
Qwen 3.8-Max has been released and currently holds the #2 position in the Vision Arena leaderboard, trailing only Claude Fable 5 Max.
Alibaba has announced the release of two new models, Qwen3.8-27B and Qwen3.8-Max.
The Qwen team has released a technical report for Qwen-UI-Agent, a real-world centric foundation GUI agent designed to operate reliably across mobile, computer-use, web, and DeepSearch environments. The system combines diverse sandbox environments with a large-scale real-device mobile runtime, interleaving GUI operations with CLI execution and supporting long-horizon tasks.
Researchers introduce Relay On-Policy Distillation (Relay-OPD) to address prefix failure in standard on-policy distillation, where student errors lead to unreliable supervision. The method uses a teacher-student continuation asymmetry as a trigger for the teacher to briefly take over and redirect the trajectory before the student resumes.
Microsoft has released Mage-VL, an efficient 4B-parameter multimodal foundation model for image and video understanding that uses a codec-native approach to streamline visual processing. The system separates video streams into anchor (I) frames and predicted (P) frames, retaining only patches where the codec allocates bits to reduce visual token consumption by over 75%.