open AI models

Nvidia is reportedly in talks to acquire Reflection AI, a US startup developing 'open' AI models. The Financial Times reported the negotiations, which would mark a significant move by the chipmaker into AI model development beyond its dominant semiconductor business. Discussion online centres on what the deal would mean for the open-model ecosystem and for competition in AI.

A new open-source project, Strata, claims to run the Qwen 3.8 Flash Next model (125B parameters) on consumer hardware like Nvidia's RTX 4090, reportedly reaching 100 tokens per second. The claim, shared on Hacker News, is drawing attention because it would bring large-model performance within reach of gaming PCs, but the speed figures have not been independently verified.

Nvidia is reportedly considering acquiring Reflection AI, an open-weight AI model startup valued at around $25 billion that Nvidia already counts among its backers. The potential deal would mark a significant step in Nvidia's push beyond chips into AI model development, and it is drawing attention across finance and technology circles as observers weigh what it would mean for competition in the open-source AI space.

Discussion is surging around the rapid growth of open-source artificial intelligence and why it matters. Supporters argue freely available models are narrowing the gap with closed systems from major labs, lowering costs for developers and raising questions about access, safety and who controls AI technology.

Nvidia may be preparing a multibillion-dollar investment in or acquisition of an open-source AI startup, in a deal reportedly worth around $25 billion. The move would deepen the chipmaker's hold on the artificial intelligence boom, extending its influence beyond hardware into open-source AI models. No target company or deal terms have been confirmed, and Nvidia has not publicly commented on the report.

Nvidia is reportedly in negotiations to acquire an open-source AI research lab that it only recently invested in. The move would deepen the chipmaker's hold on the AI software ecosystem, following its string of strategic bets on startups building models and tooling around its hardware. Details on price and timing have not been disclosed.

A developer has built an open-source Android project that takes a photo of the sky from a balcony each evening, measures it into eight colour bands, and turns the result into a named colour card. An on-device Gemma language model writes a single descriptive line for each card but is given no say in any other decision.

Nvidia is in talks to acquire Reflection AI, a startup developing open AI models, according to a Financial Times report carried by Seeking Alpha. If completed, the deal would deepen Nvidia's push into the AI software and model layer beyond its dominant chip business. Details such as price and deal status have not been confirmed by either company.

Misha Laskin, associated with Beam, has set out the case that openly available AI models will ultimately prevail over closed alternatives. The piece, titled 'Beam and the Open Frontier', frames open models as the winning path for the industry. Beyond the headline, no further details of his argument or reactions to it are available.

A developer has unveiled Touch Grass, a mobile-first gardening companion app built for Hacktoberfest 2026. The tool runs entirely on-device using Google's Gemma open-weight models, the Mastra agent framework and Ollama's local runtime, meaning user data stays private. The project is part of the Hacktoberfest open-source challenge, where contributors build and share software throughout October.

A prediction market bet on open-source AI reaching the frontier has climbed to 86%, driven by momentum from Reflection and Mistral. Traders increasingly expect open-weight models to match or rival closed frontier systems, reflecting a broader shift in the AI debate. The development fuels ongoing discussion about whether proprietary labs can keep their lead as open alternatives close the gap.

DeepSeek Harness, an open-source tool for working with DeepSeek AI models, has released version 0.2.1-alpha.2 as a pre-release. The update adds reasoning translation, letting users read the model's chain-of-thought in another language, and support for Git worktrees so developers can run parallel coding tasks in separate working directories. The release is aimed at developers using DeepSeek models for agentic coding workflows.

Alibaba's chairman has argued that open-source artificial intelligence offers Europe the best route to technological independence. His claim is that freely available models would let European companies and governments build AI capabilities without depending on American or Chinese proprietary systems. The remarks touch a live debate in Europe over digital sovereignty and regulation, with supporters seeing open-source models as a strategic alternative to US platforms.

Nvidia is reportedly considering a bigger commitment to open-weight AI, including a possible acquisition of ReflectionAI. The news, picked up by outlets including Digital Today, suggests the chipmaker could expand beyond hardware into open AI model development. Details on timing, deal size or confirmation from either company were not provided, so the scope of any transaction remains unknown.

Nvidia is reportedly in talks to acquire Reflection AI, a startup focused on open AI models. The deal, if completed, would deepen the chipmaker's push into AI software and model development, expanding beyond its dominant hardware business. Details such as price and timing have not been disclosed, and it remains uncertain whether negotiations will result in an agreement.

Nvidia is in talks to either invest further in or fully acquire Reflection AI, according to the Financial Times. Reflection AI is a US startup founded by former DeepMind researchers that is developing autonomous coding and open frontier AI models. The move would deepen Nvidia's footprint in the AI software layer beyond its dominant chip business. Details of valuation and deal terms have not been disclosed.

Anthropic has restricted Claude's ability to access the live internet after problems surfaced during evaluation. The move comes after failures were identified in how the AI model handled real-time web access, prompting the company to pull back the capability. The decision highlights ongoing safety concerns among AI developers about letting models browse the open web unsupervised, and it has drawn attention to how companies test and gate advanced features before wider release.