Alibaba released Qwen3.8-Max on August 3, 2026 as a free, open-weight 2.4 trillion parameter model that claims wins over Claude Opus 4.8 on several benchmarks and splits results with GPT-5.6 Sol, including a reported 93.0 score on PaperBench and 82.8 on IFBench. Alibaba said the weights would be published on Hugging Fa The article notes that benchmark scores come largely from Alibaba's own testing harness and should be read with care, but argues the pricing and public weights still threaten the economics of paid frontier AI by giving developers a near-frontier alternative without vendor lock-in. The coverage frames the release as a d

Alibaba's Qwen team released Qwen3.8-Max on August 3, 2026, the latest and most powerful model in the Qwen family, built on the Qwen 3.5 architecture with 2.4 trillion total parameters, 95 billion activated parameters, and a 1 million-token context window. The company said model weights would be made publicly available The Qwen3.8 API service is priced at 12 yuan ($1.7) per million input tokens and 36 yuan per million output tokens for China users, while overseas developers are charged $2 per million input tokens and $6 per million output tokens, with implicit cache-hit pricing at $0.25 per million tokens. The piece frames these move

Alibaba unveiled Qwen3.8-Max on August 3, 2026, its largest AI model to date with 2.4 trillion parameters, putting it close to Moonshot AI's Kimi K3 at 2.8 trillion parameters. The model immediately shot up AI text and visual model leaderboards after release, sending Alibaba's shares up roughly 7% in Hong Kong trade, a The Reuters report positions Qwen3.8-Max alongside DeepSeek's V4-Flash as evidence that Chinese AI firms are delivering powerful systems at ultra-low cost, with Omdia analyst Lian Jye Su noting that open-weight Chinese models meet demand for affordable, transparent, and accessible systems. The story highlights Alibaba'

Alibaba released Qwen3.8-Max on July 19, 2026 as a 2.4 trillion parameter model that became available immediately through Alibaba Cloud's Token Plan and Qoder platforms, with a 90% preview discount. The model reportedly positions itself as the runner-up to Anthropic's Fable 5, and Alibaba's broader open-source Qwen fam Alibaba owns about 36% of Moonshot AI following an ~$800 million investment in February 2024, meaning Qwen3.8-Max and Moonshot's 2.8 trillion parameter Kimi K3 represent both internal and external competition. Alibaba's shares gained over 3% in pre-market trading following the announcement, with both firms betting that

On July 19, Alibaba Cloud released the Qwen3.8-Max preview, a 2.4 trillion parameter native multimodal Mixture-of-Experts model and Alibaba's first to surpass the trillion-parameter threshold, with an explicit commitment to return to an open-source roadmap. Reported gains include 22.8% in coding and 44.4% in office pro Alibaba is using a 90% limited-time discount and a daily-update model strategy to gather user feedback, and Tongyi Lab has undergone three organizational restructurings in the 140 days since former technical lead Lin Junyang's departure, pivoting toward a commercially driven Token Factory model. With open-source models

An infrastructure-focused analysis of Qwen3.8-Max-Preview notes that as of July 22, 2026, Alibaba had published no benchmark table, no model card on Hugging Face, no license, no open-weight date, and no active-parameter count, despite the preview going live. The model is callable through Alibaba's Token Plan subscripti The analysis argues the headline 2.4T number is the wrong planning figure: in a sparse MoE, total parameters set the memory floor while active parameters set the compute bill, and Alibaba withheld the latter, making external cost modeling impossible. By contrast, DeepSeek V4 Pro (1.6T total, 49B active) and GLM-5.2 (74

Alibaba shares surged more than 3.5% in overnight trading on July 20, 2026, after the Chinese giant unveiled Qwen3.8-Max in preview, described as "second only to Fable 5." With 2.4 trillion parameters, the model is expected to go open-weight soon and is available to try on Alibaba's Token Plan, Qoder, and QoderWork sur Across cited benchmarks, Kimi K3 reportedly trailed only OpenAI's GPT-5.6 Sol and Anthropic's Fable 5, while outperforming them on certain coding and general agent tasks at a cheaper price. The latest Chinese AI launches have pressured competitor Zhipu's shares down more than 45% in five days, while MiniMax lost nearly

Alibaba previewed Qwen3.8-Max-Preview on July 19, 2026, at the World Artificial Intelligence Conference in Shanghai, unveiling a sparse mixture-of-experts model with 2.4 trillion total parameters. The Qwen team said on X that the model ranks "second only to Fable 5," a reference to Anthropic's recent release, though th The preview is accessible through Alibaba's Token Plan subscription and its Qoder and QoderWork developer platforms, priced at 10% of the anticipated standard rate during the trial window. The launch comes days after Moonshot AI's Kimi K3 (2.8T parameters) set the record for the largest open-weight model, directly posi