New Open Model Elevates AI Images, Ends 'AI Look'
The landscape of artificial intelligence continues its rapid evolution, with new developments emerging across image generation, infrastructure expansion, and market dynamics. A newly released open model promises to further blur the lines between AI-generated images and reality, while OpenAI expands its ambitious data center project into Europe, and a recent report reveals a significant shift in enterprise AI spending.
Redefining AI Image Realism
AI image startup Black Forest Labs (BFL) and creative platform Krea have unveiled FLUX.1 Krea, an open-weight image model designed to overcome the common “AI look” characterized by unnatural textures, blurry backgrounds, and oversaturated colors. The model was trained on a meticulously curated and diverse dataset specifically to avoid these typical AI outputs, aiming for enhanced photorealism and quality.
FLUX.1 Krea is touted by its creators as state-of-the-art among open models, demonstrating performance that rivals leading closed systems—including BFL’s own FLUX 1.1 Pro—in human preference tests. This release is fully compatible with the FLUX.1 developer ecosystem, facilitating easy integration for developers and into various applications. The availability of such a high-quality open model signifies a major leap in AI image generation, suggesting that the aesthetic shortcomings previously associated with AI can be largely resolved through targeted training. This advancement is expected to make it even more challenging for observers to distinguish between real human photographs and AI-generated imagery.
OpenAI Expands Stargate Initiative to Europe
OpenAI has announced the launch of Stargate Norway, marking the company’s first major AI data center in Europe. Situated near Narvik, the facility is projected to house 100,000 Nvidia GPUs and operate entirely on renewable energy by late 2026.
This ambitious project will commence with 230 megawatts (MW) of capacity, with plans for expansion to 520MW, positioning it as one of Europe’s largest AI computing centers. The initiative strategically leverages Norway’s naturally cool climate and robust renewable energy grid, with a notable feature being the redirection of waste heat from the GPUs to power local businesses. Norwegian industrial conglomerate Aker and infrastructure firm Nscale have committed $1 billion to the initial phase, establishing a 50/50 ownership split. Furthermore, Norway becomes the first European nation to partner with OpenAI under its “OpenAI for Countries” program, introduced in May. This expansion into Europe, alongside recent significant deals, paints a picture of aggressive global scaling for OpenAI’s infrastructure, countering earlier speculation of potential setbacks for the Stargate project.
Enterprise AI Market Sees Significant Shifts and Spending Surge
A mid-year Large Language Model (LLM) market report from Menlo Ventures indicates a substantial surge in enterprise AI spending, with Anthropic emerging as the new market leader in model usage. The report, based on a survey of 150 technical leaders, found that enterprise LLM API spending doubled to $8.4 billion in the past six months alone.
Anthropic now holds the top spot with a 32% market share, surpassing OpenAI (25%) and Google (20%). This represents a significant shift from 2023, when OpenAI commanded a dominant 50% share. The report also highlighted code generation as AI’s “breakout use case,” with developers increasingly adopting an ecosystem of AI coding agents and integrated development environments (IDEs) rather than single-product tools. A key finding is the high “stickiness” of AI platforms: enterprises rarely switch providers once adopted, with 66% opting to upgrade models within their existing ecosystem rather than changing vendors. Additionally, the report noted a stagnation in open-source LLM usage among enterprises, which are prioritizing performance and reliability over cost. This trend suggests the emergence of a “winner-take-most” market, where the early technological advantages of leading AI providers are solidifying into significant competitive barriers.
Other AI Developments
In related news:
Cohere introduced Command A Vision, a new model achieving state-of-the-art performance in multimodal vision tasks tailored for enterprises.
OpenAI has reportedly reached an annualized revenue of $12 billion for 2025, with its ChatGPT platform attracting approximately 700 million weekly active users.
StepFun released Step3, an open-source multimodal reasoning model that demonstrates high performance at low cost, outperforming models like Kimi K2, Qwen3, and Llama 4 Maverick.
According to a new report, video model developers Runway and Luma AI are exploring robotics training and simulations as a potential revenue stream.
AI infrastructure platform Fal secured a new $125 million funding round, elevating the company’s valuation to $1.5 billion.
Agentic AI startup Manus launched Wide Research, a feature that leverages agent-to-agent collaboration, enabling hundreds of subagents to collectively address a single task.