Submit a toolJoin free

Today in AI

Alibaba unveiled Qwen3.8-Max, a 2.4-trillion-parameter model for text, images and video. DeepSeek’s V4-Flash lowered the running cost of a well-known AI model in benchmark tests. AI leaders split over whether powerful models should stay open or face tighter controls. Sam Altman said AI development may need pacing so society can adapt to rising capabilities. ArcelorMittal expanded its Microsoft partnership around cloud, data governance and AI.

The latest signal

Top AI News

Five news stories from today’s digest.

AIModels

Alibaba unveils Qwen3.8-Max, its largest AI model yet

Reuters•3 Aug 2026

News Digest

Alibaba unveiled Qwen3.8-Max, describing it as the largest and most capable model in the Qwen family. Its 2.4 trillion parameters use a mixture-of-experts design that activates only part of the network for each request. The model handles text, images and video with a one-million-token context window. Independent testing must still establish its speed, cost and reliability.

Key takeaways

03
  1. Alibaba unveiled Qwen3.8-Max with 2.4 trillion parameters, making it the largest and most capable model in the company’s Qwen family overall.
  2. The mixture-of-experts design activates only part of the network for each request, pairing very large capacity with manageable inference.
  3. Alibaba will distribute the model through Model Studio, giving developers access while independent tests compare speed, cost and reliability.

Why it matters

Model selection is no longer about parameter count alone. Businesses must weigh context length, multimodal support, inference efficiency and developer access. If Qwen3.8-Max converts its scale into reliable performance at an acceptable cost, competing providers will face pressure to improve capability without allowing operating expenses to rise as quickly.

Read original story
AIModels

DeepSeek’s V4-Flash undercuts leading models on running cost

Reuters•3 Aug 2026

News Digest

DeepSeek’s V4-Flash was the cheapest well-known model to run in benchmark tests assessed by Artificial Analysis. The firm estimated an average test cost of three cents and ranked it level with Gemini 3.6 Flash, but below leading OpenAI and Anthropic systems. Its advantage lies in operating efficiency rather than overall benchmark leadership, favouring high-volume workloads.

Key takeaways

03
  1. Artificial Analysis found DeepSeek’s V4-Flash was the cheapest well-known model to run in its benchmark tests, averaging three cents per test.
  2. DeepSeek charges $0.14 per million input tokens and $0.28 per million output tokens, making small unit savings material at production scale.
  3. Its Intelligence Index score matched Gemini 3.6 Flash but trailed leading OpenAI and Anthropic systems, making cost its strongest advantage.

Why it matters

Developers have another reason to match models to individual workloads instead of defaulting to one frontier system. DeepSeek’s pricing also forces premium providers to show that stronger reasoning, reliability or tooling justifies a higher operating cost.

Read original story
AIPolicy

AI leaders divide over opening powerful models

Axios•2 Aug 2026

News Digest

AI companies are taking different positions on whether powerful models should be opened widely or contained more tightly. Nvidia and Meta favour broader access to open-weight systems, while Anthropic backs mandatory safety tests and tighter controls on advanced chips and model distillation. OpenAI supports open infrastructure but has discussed a government-controlled emergency brake.

Key takeaways

03
  1. Nvidia and Meta favour wider availability of open-weight systems, arguing that independent deployment supports research and market competition.
  2. Anthropic supports mandatory safety tests and tighter controls around advanced chips and model distillation as capabilities become more powerful.
  3. OpenAI backs open infrastructure but has discussed a government-controlled emergency mechanism that could slow development if serious risks emerge.

Why it matters

The policy outcome will shape who can build on frontier models, what compliance obligations apply and how much control governments retain after release. Businesses may gain broader access, but they could also face greater uncertainty around platform rules and deployment risk.

Read original story
AISafety

Sam Altman says AI development may need pacing

TechCrunch•2 Aug 2026

News Digest

OpenAI CEO Sam Altman said AI development may need to be paced so society can adapt to rising capabilities, without calling for a pause. TechCrunch linked the debate to recent autonomous-agent security incidents and argued that containment, testing, permissions and deployment controls matter alongside speed. Release timing and staged roll-outs may gain a larger place in AI policy.

Key takeaways

03
  1. Sam Altman said advanced AI development may need pacing so society can adapt, while stopping short of supporting a broad halt in progress.
  2. Recent autonomous-agent incidents increased scrutiny of containment, testing environments and permissions around powerful systems before wider deployment.
  3. Staged releases could become a larger part of AI policy, giving businesses more predictability while delaying access to some new capabilities.

Why it matters

Release timing and staged deployment may become more prominent in AI policy alongside conventional safety tests. Businesses could gain greater predictability from controlled launches, although access to useful capabilities may arrive more slowly.

Read original story
AIBusiness

ArcelorMittal expands Microsoft partnership around data and AI

Reuters•3 Aug 2026

News Digest

ArcelorMittal expanded its Microsoft partnership around a Cloud First, Data Centric strategy, keeping Azure as its primary cloud platform. The steelmaker will integrate Microsoft Fabric for data and analytics, Purview for governance and compliance, and Foundry for AI development. The programme also targets stronger cybersecurity and less dependence on legacy systems.

Key takeaways

03
  1. ArcelorMittal expanded its Microsoft partnership, retaining Azure as the primary platform for its cloud, data and AI modernisation programme.
  2. Microsoft Fabric will support data and analytics, Purview will cover governance, and Foundry will provide tools for AI development and testing.
  3. Financial terms were not disclosed, leaving the strategic direction clear but preventing an assessment of the partnership’s commercial scale.

Why it matters

Large organisations often need standardised data, governance and cloud foundations before AI can move beyond isolated pilots. ArcelorMittal’s approach shows why enterprise adoption is increasingly a platform decision, not simply the purchase of one AI application.

Read original story