Alibaba has unveiled Qwen3.8-Max, its newest and most powerful artificial intelligence model, marking another significant step in China’s effort to narrow the technological gap with leading U.S. AI companies and strengthen its position in the global race for frontier AI.

The model, which is scheduled for release next week, features 2.4 trillion parameters, making it the largest and most capable model in Alibaba’s Qwen family to date. Parameters are the internal numerical values learned during training that influence how an AI model understands language, identifies patterns and generates responses.

While parameter count is no longer the sole measure of AI capability, models at this scale typically possess greater reasoning capacity, stronger contextual understanding and broader task versatility when combined with advanced training techniques.

Alibaba’s announcement was well received by investors. Its U.S.-listed shares rose 4.5% in premarket trading, while its Hong Kong-listed shares gained 7%, reflecting growing confidence that the company remains one of China’s strongest contenders in generative AI.

Over the past year, Chinese developers have dramatically accelerated model releases, responding not only to intense domestic competition but also to rapid advances by U.S. companies including OpenAI, Anthropic, Google and xAI. The competitive landscape has shifted beyond chatbot performance to enterprise deployment, autonomous AI agents, multimodal reasoning and long-context processing, areas increasingly viewed as the next frontier of commercial AI.

A defining feature of Qwen3.8-Max is its one-million-token context window, allowing the model to process and reason across thousands of pages of documents or vast collections of multimedia content in a single session. That capability significantly expands the range of enterprise applications. Instead of analyzing isolated documents, the model can review entire legal case files, corporate archives, financial disclosures, scientific literature, or extensive technical documentation while maintaining contextual consistency throughout.

Long-context models are becoming more valuable for industries such as finance, healthcare, engineering, legal services and scientific research, where professionals routinely work with large datasets that exceed the limits of conventional AI systems.

Alibaba said the model is designed to support software engineering, research, coding, workplace productivity, visual intelligence and other complex, long-duration tasks.

Perhaps the most notable aspect of the announcement is Alibaba’s emphasis on autonomous execution rather than conventional chatbot interactions.

According to the company, Qwen3.8-Max demonstrated the ability to operate with minimal human intervention during internal testing.

“In one internal test, it spent 16 days building and improving an AI coding tool by writing code, testing it, fixing errors, and refining its work on its own,” Alibaba said.

That reflects one of the most important shifts occurring across the AI industry. Rather than simply answering prompts, leading AI developers are racing to build intelligent agents capable of independently planning projects, executing multi-step assignments, evaluating their own output and continuously improving results over days or even weeks.

If those capabilities prove reliable in commercial environments, AI could increasingly automate knowledge-intensive work traditionally performed by software engineers, researchers, analysts and consultants.

Enterprise Adoption Becomes The Next Battleground

Alibaba also highlighted the model’s ability to perform practical business tasks, including reviewing legal documents, conducting financial analysis and supporting architectural 3D modelling. That shows that competition among AI companies is now centered on enterprise productivity rather than consumer chatbots alone.

Businesses represent one of the largest long-term revenue opportunities for AI developers because enterprise customers are typically willing to pay premium subscription fees for models that improve productivity, reduce labor costs and accelerate decision-making. This shift mirrors recent moves by OpenAI, Microsoft, Anthropic and Google, all of which have expanded aggressively into enterprise AI software and workplace automation.

Alibaba said Qwen3.8-Max also represents a significant advance in multimodal AI. The model can understand hundreds of pages of documents, entire television series, or more than 100 hours of livestream video before transforming that information into searchable and interactive knowledge repositories.

Such capabilities demonstrate how AI models are evolving beyond text generation into comprehensive information-processing systems capable of analyzing text, images, and video simultaneously. Multimodal reasoning is widely regarded as one of the industry’s next major growth areas because businesses require AI systems capable of interpreting multiple forms of data within a single workflow.

Alibaba also released benchmark results comparing Qwen3.8-Max with leading global AI models. According to the company, the model achieved performance comparable to, and in some cases surpassing, Anthropic’s Fable 5 across several evaluation benchmarks. Alibaba said Qwen3.8-Max ranked second only to Fable 5 in Vision Arena and fifth in Text Arena.

Although benchmark results should be interpreted cautiously because methodologies vary and companies often highlight favorable tests, the findings nevertheless suggest Chinese AI developers continue to reduce the performance gap separating them from leading U.S. models.

China’s AI Competition Intensifies

The announcement comes only weeks after Moonshot AI introduced Kimi K3, a 2.8-trillion-parameter model that currently ranks as China’s largest AI model by parameter count.

The rapid succession of capable models shows that inside China’s AI ecosystem, competition is becoming as intense as it is outside. Companies including Alibaba, Moonshot AI, Tencent, Baidu and DeepSeek are competing aggressively to establish technological leadership while simultaneously expanding their enterprise customer base.

The pace of development has been particularly notable given continuing U.S. restrictions on exports of advanced AI chips to China. Rather than slowing innovation, those restrictions have encouraged Chinese developers to optimize model architectures, improve software efficiency, and maximize performance using available computing resources.

Beyond the technical specifications, Qwen3.8-Max indicates that, in the global AI race, the competition is no longer defined solely by who builds the largest chatbot. It now revolves around which companies can deliver AI systems capable of functioning as autonomous digital workers that generate measurable economic value across industries.

The European Union has entered a new phase of artificial intelligence regulation after granting its executive arm sweeping enforcement powers over advanced AI models, significantly expanding its authority to inspect frontier systems, restrict market access and impose multimillion-euro penalties on developers that fail to comply with the bloc’s landmark AI legislation.

The new powers, which took effect on Sunday under the EU AI Act, strengthen the authority of the European Commission and its AI Office, placing leading U.S. developers including OpenAI, Anthropic and Google under a far more rigorous regulatory regime if they offer general-purpose AI models within the European market.

The measures represent one of the world’s most comprehensive attempts to regulate frontier AI and reinforce the EU’s ambition to become the global standard-setter for artificial intelligence governance.

Under the new framework, the European Commission can require developers of general-purpose AI (GPAI) models to submit their systems for evaluation before they are made publicly available in the European Union. Regulators may also demand technical information about models, conduct compliance assessments and, where necessary, limit or prohibit access to the EU market if they conclude a model presents unacceptable risks or violates the AI Act.

Companies that fail to comply face penalties of up to €15 million ($17.3 million) or 3% of annual global turnover, whichever is higher.

Importantly, the Commission can also impose fines for procedural violations, including refusing information requests, providing misleading responses or obstructing regulatory evaluations, even if no underlying safety breach is established.

Legal experts say those provisions substantially expand regulators’ leverage over AI developers.

Elisabetta Righini, a partner at Sidley Austin, said the rules apply to any company offering a general-purpose AI model in Europe, regardless of where it is headquartered.

“A U.S. address does not put a lab outside the EU regulator’s reach,” she said, noting that non-European providers must appoint an EU-based authorized representative to serve as their regulatory point of contact.

AI Act Enters Enforcement Phase

The expanded supervisory powers form part of the phased implementation of the 2024 EU AI Act, the world’s first comprehensive legal framework governing artificial intelligence. Rather than taking effect all at once, the legislation is being introduced gradually, with different obligations becoming enforceable over several years.

The latest phase focuses on general-purpose AI models, particularly frontier systems capable of performing a broad range of tasks and posing systemic risks because of their scale and capabilities.

Henna Virkkunen, Executive Vice-President for Tech Sovereignty, Security and Democracy at the European Commission, said stronger oversight is necessary as AI systems become increasingly powerful.

“Harms can occur if AI is not properly designed and used and the most advanced models create risks on an entirely new scale,” she said.

Stoking Friction with U.S. AI companies

The new enforcement powers are likely to intensify tensions between Brussels and major American AI developers, many of whom have questioned aspects of the EU’s regulatory approach.

Recent disputes show that the relationship remains complex.

Reuters reported last week that the European Union has been in discussions with OpenAI and Anthropic following alleged cyber incidents involving their AI models. OpenAI confirmed it has been engaging with the EU AI Office. The two companies have also faced growing pressure from European regulators seeking greater transparency into the capabilities and risks of their frontier AI systems.

In May, European authorities reportedly spent months seeking access to Anthropic’s Mythos model before the company agreed to cooperate. Tom Gordon, OpenAI’s Vice President for Europe, the Middle East and Africa, said the company has worked closely with European policymakers during implementation of the AI Act.

“We’ve collaborated closely with the European Commission and the wider ecosystem on implementing the AI Act, including its Codes of Practice, and will continue working together to help Europe realize the benefits of the Intelligence Age,” Gordon said.

Google likewise said it remains committed to complying with the new regulatory framework while continuing to invest in European AI infrastructure and innovation.

The tougher AI oversight also arrives against the backdrop of widening trade and regulatory disputes between Washington and Brussels. The European Union has been pursuing a strategy centered on technological sovereignty, seeking to reduce its dependence on U.S.-based digital platforms, cloud infrastructure and artificial intelligence providers.

That strategy has frequently brought European regulators into conflict with major American technology companies.

Last month, Google was fined $1 billion after EU regulators concluded the company had unfairly favored its own services. The decision prompted President Donald Trump to threaten the bloc with substantial tariffs, highlighting the growing overlap between technology regulation and international trade policy.

The Commission has also demonstrated a willingness to scrutinize AI-enabled products beyond traditional language models.

In January, European regulators opened an investigation into X over the dissemination of sexually explicit content generated by its AI chatbot, Grok.

Although the AI Act is European legislation, its influence extends far beyond the bloc. Because multinational AI developers generally operate unified global platforms, many analysts expect compliance measures adopted for Europe to shape product development worldwide.

The law also reinforces the “Brussels Effect,” whereby companies frequently adopt European regulatory standards across global operations rather than maintaining separate compliance systems for different jurisdictions.

For AI developers, compliance is becoming a strategic issue alongside model performance and commercialization. Companies must now demonstrate not only technological capability but also transparency, risk management and regulatory readiness as governments around the world move to establish formal oversight of powerful AI systems.

The AI Act uses a risk-based approach, imposing progressively stricter obligations depending on the potential societal impact of AI systems. General-purpose AI models, particularly frontier systems capable of broad reasoning and autonomous decision-making, are subject to enhanced transparency, documentation and safety requirements.

The latest enforcement powers mark a significant milestone in the Act’s implementation, giving the European Commission practical authority to investigate, evaluate and sanction AI providers.

OpenAI has established an early lead in the race for influence on Capitol Hill, accounting for nearly 90% of identifiable artificial intelligence spending by the U.S. House of Representatives, making ChatGPT the default generative AI platform for lawmakers and congressional staff.

A CNBC analysis of House disbursement records shows OpenAI captured roughly nine out of every 10 dollars spent on named AI tools between April 1, 2025 and March 31, 2026. The spending records, while incomplete, offer one of the clearest public snapshots yet of how generative AI is being integrated into the daily operations of Congress.

The analysis identified at least $113,740 spent on identifiable AI vendors during the period. OpenAI accounted for $100,580 across 798 transactions, representing approximately 88% of total spending and 96% of all AI-related purchases. Rival Anthropic ranked a distant second, with Claude generating $13,160 from just 37 transactions.

The figures likely understate AI adoption across Congress because they exclude free chatbot accounts, AI capabilities bundled into broader software subscriptions such as Microsoft Copilot, reimbursements that do not identify the underlying software, and nearly all Senate usage. Even so, the data indicates that at least one-sixth of House offices have already purchased dedicated AI tools, suggesting that generative AI is rapidly becoming embedded in legislative work.

The spending race is strategically significant because the companies competing for congressional adoption are the same firms lobbying lawmakers and regulators over AI policy, procurement rules and future regulation.

Congressional offices are increasingly relying on AI to summarize legislation, prepare policy briefings, draft constituent correspondence, produce hearing materials, analyze bills and organize research, allowing staff to process growing workloads with limited resources.

Representative Julie Fedorchak, a Republican from North Dakota, said ChatGPT has transformed how her office prepares legislative materials.

“My staff used to spend hours writing memos to prepare me for the dozen-plus meetings I would have in a day,” Fedorchak said.

She said her office created a searchable database containing legislation, constituent submissions and policy materials that now enables ChatGPT to generate briefing documents automatically.

“It’s saving dozens and dozens of hours every week of staff time,” she said.

The productivity gains come as congressional offices face mounting pressure from growing volumes of information, a trend that AI itself is accelerating.

Marci Harris, chief executive of the POPVOX Foundation, said staff are confronting an unprecedented information overload.

“Staffers are famously overwhelmed, and actually AI is exacerbating that,” Harris said, noting that lobbyists, advocacy groups and constituents are themselves using AI to produce longer, more detailed and more frequent communications directed at Congress.

While OpenAI enjoys a commanding lead, Harris cautioned that Congress should avoid becoming dependent on a single platform as lawmakers prepare legislation governing the industry.

“It’s going to be their job to regulate it, so they need to be able to get their hands on all of them and understand the way that they work,” she said.

The political breakdown of AI spending reveals another notable trend. Democratic House offices accounted for $54,165 in identified AI purchases, more than three times the $15,782 spent by Republican offices.

ChatGPT subscriptions appeared in 44 Democratic offices and 27 Republican offices, while Anthropic’s Claude appeared in five Democratic offices and just one Republican office.

The figures are striking because many Democratic lawmakers have simultaneously raised concerns about AI’s potential impact on employment, privacy, elections, misinformation, civil rights and market concentration.

Congressional officials stress that AI is intended to assist staff rather than replace human judgment.

Representative James Walkinshaw, a Democrat from Virginia, said AI can streamline administrative work but should never make decisions affecting constituents.

“Our caseworkers in our district office can use AI to help organize casework requests, maybe expedite them, move things along faster,” Walkinshaw said.

“But when they’re making decisions related to a constituent’s case — are we going to advocate for our constituent with the Social Security Administration or with IRS? — that has to be a human decision, not an AI decision.”

OpenAI’s dominance appears to stem from its early entry into Congress.

In 2023, the House Digital Service launched an AI working group and distributed the first 40 ChatGPT Plus licenses to bipartisan congressional staff, giving OpenAI a significant first-mover advantage before competing products gained traction.

Marci Harris compared ChatGPT’s early brand recognition to household consumer products.

“OpenAI was the firm that they heard about,” she said.

“In the early days, ChatGPT was almost synonymous with AI — the way Kleenex is with tissues or Band-Aid is with bandages.”

The company also strengthened its presence through training initiatives. The Congressional Management Foundation partnered with OpenAI to provide ChatGPT training sessions for House and Senate staff, with OpenAI engineers participating directly in the programme. Some congressional offices have also received product training from OpenAI and Microsoft.

Meanwhile, AI companies are intensifying their lobbying efforts in Washington as Congress and federal agencies determine how the technology should be regulated.

According to Issue One, OpenAI spent $1 million on federal lobbying during the first quarter of 2026, an 82.5% increase from a year earlier. Anthropic spent nearly $1.6 million during the same period, more than tripling its lobbying expenditure year-on-year.

Interest in Anthropic also rose after a high-profile dispute with the U.S. Department of Defense earlier this year.

“After the whole Department of Defense-Anthropic blowup, we had people calling us saying, ‘I want to use Claude. Tell me about Claude,'” Harris said.

The concentration of congressional AI spending has also prompted governance questions.

Eric Petry of the Brennan Center for Justice said government institutions should ensure AI purchasing decisions remain driven by performance rather than politics.

“It may be the case that one product is better suited for the government applications, or maybe one product is better than the other,” Petry said.

“But if it’s based on partisan preferences, or there’s politics at play, then it becomes really concerning.”

He added that the same principle should apply whether the government is purchasing enterprise-scale AI systems or individual software subscriptions.

“We want to make sure that taxpayer dollars are being used to give the government the best tools available, and not to reward political favorites,” Petry said.

Although the available spending records represent only a partial picture of AI adoption across Congress, they illustrate OpenAI’s significant first-mover advantage at a time when lawmakers are shaping the regulatory framework that will govern the industry’s future. The data also suggests that ChatGPT has evolved beyond a productivity tool into core legislative infrastructure, positioning OpenAI with substantial influence as artificial intelligence becomes increasingly embedded in policymaking on Capitol Hill.