SShortSingh.
Back to feed

AMD Invests $5B in Anthropic as Microsoft Fine-Tunes Model Built on Alibaba's Qwen

0
·1 views

AMD has committed up to $5 billion in Anthropic, while Microsoft simultaneously expanded its partnership with French AI lab Mistral to strengthen its foothold in European sovereign AI markets. Microsoft also released Fara1.5-27B, a vision-based browser agent fine-tuned from Alibaba's Chinese-developed Qwen3.5-27B model, underscoring Western reliance on Chinese base models despite rising trade tensions. Austria launched a sovereign federal AI platform called GovGPT, built on Mistral and hosted entirely on domestic infrastructure, serving approximately 180,000 government employees. The US government accused Chinese lab Moonshot AI of covertly distilling capabilities from Anthropic's models to accelerate its Kimi K3 release, allegedly using export-controlled Nvidia GPUs. These developments collectively signal that the global AI race is increasingly interdependent, with Western, Chinese, and European actors routinely building on each other's research despite geopolitical rivalries.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Developer Rebuilds Personal Site Using Astro, Bun, and Cloudflare for Better Performance

A developer has rebuilt their personal website, replacing a heavier Docusaurus and React setup with a lighter stack based on Astro, Bun, and Cloudflare Pages. The new architecture delivers under 10 KB of JavaScript on most pages and achieves strong Core Web Vitals scores. Key features include smart image optimization, internationalization via content collections, and dark mode without layout shifts. Deployment is automated through a GitHub Actions and Cloudflare Pages pipeline. Local AI tools running on the developer's Mac also played a role in shaping the rebuild process.

0
ProgrammingDEV Community ·

USB 3.2 Gen 2 Hubs Offer 10Gbps Shared, Not Per-Port, Bandwidth

A USB 3.2 Gen 2 hub provides a single 10Gbps upstream link to the host, meaning all connected devices share that total bandwidth rather than each receiving a dedicated 10Gbps channel. Real-world throughput is further limited by the host port, cables, hub controller, peripherals, and operating system, with the connection defaulting to the slowest compatible component in the chain. The USB Implementers Forum now recommends calling this standard 'USB 10Gbps' to reduce confusion with the separate 20Gbps Gen 2x2 specification. Users running multiple high-bandwidth devices simultaneously — such as two fast external SSDs — should note that combined traffic must fit within the single upstream link, making a USB4 or Thunderbolt dock a better choice for demanding workloads. When purchasing, buyers should verify the numeric data rate on product spec sheets rather than relying on connector type or generation labels alone.

0
ProgrammingDEV Community ·

Meta's Llama 4 Maverick API Tested for Production Readiness

Meta's Llama 4 Maverick is an open large language model being evaluated for real-world production use via its API. Developers have been conducting hands-on testing to assess its performance, reliability, and practical capabilities. The evaluation focuses on whether the model meets the demands of production-grade applications. As an open model, Llama 4 Maverick represents Meta's continued push into the competitive AI landscape.

0
ProgrammingDEV Community ·

Agent-Up lets multiple coding agents share an app without full Docker containerization

Running multiple coding agents on the same application typically leads developers to containerize everything in Docker, but this approach introduces friction through stale images, slow rebuilds, and added debugging complexity. Agent-Up is an open-source tool that addresses the core collision points — ports, processes, browser profiles, logs, and environment state — without requiring full app containerization. Each agent workspace receives its own isolated port range, allowing apps to run normally via standard commands like npm run dev or dotnet run. The tool follows a split model: application code runs locally while infrastructure components such as databases, queues, and caches remain in Docker. Agent-Up coordinates both layers as a single unified workspace, complementing Git for source isolation and Docker for infrastructure isolation.

AMD Invests $5B in Anthropic as Microsoft Fine-Tunes Model Built on Alibaba's Qwen · ShortSingh