AI-generated images have made amazing leaps recently — from simple cats and flowers to complex scenes with humans, horses, and intricate text layouts. But if you’ve spent any time with AI art, you probably noticed a recurring theme: despite technical prowess, many images still carry that unmistakable “AI look”. You know, those blurry backgrounds, soft textures, and somewhat dull or waxy skin that feels just a bit off. I recently discovered that the team behind FLUX.1 Krea is tackling exactly this problem with a fresh, unapologetically opinionated approach that’s worth digging into.

Beyond benchmarks: When AI image quality means more than just metrics

It turns out that the usual way we measure image model success—like checking if the AI got the prompt right or scored well on benchmarks like FID or CLIP—is only part of the story. According to recent insights, these standard benchmarks often miss what users truly want: images that feel authentic, stylistically diverse, and creatively engaging without screaming “made by AI.”

In fact, many popular aesthetic scorers and filters, like LAION-Aesthetics, tend to favor certain biased traits like bright images or soft textures. This means training a model on such scores can inadvertently bake in those very biases and reinforce the “AI look” rather than eliminate it.

The FLUX.1 Krea team recognized this mismatch and decided to focus on what really matters: delivering AI art that doesn’t look AI-generated. This means reevaluating their training data, metrics, and model architecture through a lens that values true aesthetic quality over just prompt adherence or simplistic scoring.

Pre-training vs post-training: The art of mode coverage and mode collapse

One of the most enlightening parts I came across was how FLUX.1 Krea approaches training in two distinct but complementary phases:

• Pre-training: Maximize diversity and mode coverage of the visual world. The model learns everything from objects and styles to places and people, absorbing both good and “bad” examples so it knows what to avoid later.
• Post-training: Carefully sculpt and bias the model towards desirable aesthetic modes by “collapsing” undesired outputs. This stage fine-tunes the model towards the opinionated aesthetic vision of the creators.

This perspective reminded me of Michelangelo’s quote that “the sculpture is already complete within the marble block” — the goal here is to chisel away the superfluous parts and reveal the desired form inside.

Interestingly, the FLUX.1 team needed a “raw” base model that was not already heavily finetuned or baked into a certain style. They partnered with Black Forest Labs to get flux-dev-raw, a 12B parameter diffusion transformer model that knew the world well but was still malleable enough to shape.

Opinionated aesthetics: Why mixing tastes can water down AI art

Here’s where things get really interesting. The team found that trying to please everyone by training on broad preference datasets led to images that were, ironically, less satisfying—too symmetric, too soft, and drifting back towards the dreaded “AI look.”

Turns out that aesthetics are deeply personal and subjective. Trying to blend multiple tastes ends up producing a bland “average” that nobody really loves. Instead, the FLUX.1 team took a bold stance: align the model strongly with a clear, specific aesthetic direction that reflects their own artistic preferences.

This approach means that for users who want to explore vastly different styles—like high fashion photography versus minimalism—prompting alone might not cut it. Many turn to add-on techniques like LoRAs for style control. The FLUX.1 strategy embraces the idea that a model overfitting to a well-defined style can actually be a feature, producing better initial outputs requiring less tinkering.

What really moved the needle: quality over quantity and human feedback

When it comes to post-training data, the team discovered that having a smaller, carefully hand-curated dataset of less than a million images beats massive generic datasets. Their preference labels—even simple pairs of images rated on aesthetics—were gathered thoughtfully, focusing on strict style consistency and knowledgeable annotators deeply aware of the model’s flaws.

They then used a combination of supervised finetuning and a unique reinforcement learning from human feedback (RLHF) approach called TPO (a variant of preference optimization) to further push the model’s alignment with their aesthetic goals. Multiple rounds of this fine-tuning helped the model nail not just image quality but the feel and style of their desired look.

Looking ahead: Personalized AI art and broader creative horizons

FLUX.1 Krea is just the starting point for a bigger vision. The plan is to keep improving the core capabilities and expand into new visual domains for richer creativity. But perhaps the most exciting direction is aesthetics personalization—building models that tailor outputs to individuals’ unique tastes and preferences.

Imagine a future where your AI art tool understands exactly what style and nuances you want, going beyond general opinionated aesthetics to something truly personal and expressive. The journey of FLUX.1 Krea reveals how foundational this fine balance between technical prowess, data curation, and personal artistic vision is.

Key takeaways

• Classic AI image benchmarks don’t always capture what users want: authentic, creative, non-AI-looking art.
• Training in two phases—pre-training for diversity, post-training for focused aesthetic bias—helps achieve superior results.
• Opinionated model training tailored to a specific artistic vision often outperforms trying to please everyone simultaneously.
• High-quality, carefully curated datasets paired with human feedback can dramatically improve final model aesthetics even with less data.

Wrapping up

Diving into the development story of FLUX.1 Krea gave me a refreshing perspective on how advanced AI image models can go beyond just technical feats to truly meet creative desires. The team’s willingness to challenge norms—whether by questioning typical benchmarks or embracing a strong aesthetic opinion—shows a maturity in generative AI development that’s needed for the field to progress.

For anyone exploring AI-generated artwork, FLUX.1 Krea offers a promising step toward images that not only impress with detail and accuracy but also feel genuinely artistic and alive. I’m excited to see how this open model will inspire the community and what new styles and applications will emerge as AI continues to get smarter and more personally expressive.

As they say, making AI images that don’t look like AI is no small feat—but FLUX.1 Krea shows it’s definitely possible.

The post The end of the ‘AI Look’: Krea’s FLUX.1 delivers true photorealism appeared first on Aiholics: Your Source for AI News and Trends.

If you’ve been tracking the world of large language models (LLMs) and generative AI, you’ve probably noticed the ground shifting beneath our feet, especially in enterprise adoption. I recently came across some fascinating insights that reveal a major shakeup in the LLM market halfway through 2025.

Here’s the scoop: while OpenAI once dominated enterprise usage, it’s now been overtaken by Anthropic, according to a report from Menlo Ventures. This shift signals not only a change in market leadership but also highlights evolving priorities around model capabilities, cost dynamics, and the emergence of what’s being called the “year of agents.” Let’s unpack what’s really going on.

Anthropic’s meteoric rise: why this newcomer is winning the AI race

Not long ago, OpenAI controlled about half of enterprise LLM usage. Fast forward to mid-2025, and that share has shrunk to roughly a quarter. Meanwhile, Anthropic has surged ahead, claiming about 32% of enterprise usage, surpassing OpenAI and even Google.

What powered Anthropic’s rise? It boils down to a few key breakthroughs centered on their Claude model series—especially Claude Sonnet 3.5, 3.7, and the latest Claude Sonnet 4.

• Code generation is the first real killer app for AI. Claude quickly became a favorite among developers, capturing 42% of the market — twice the share of OpenAI’s models. This alone turned code generation from a niche product into a $1.9 billion ecosystem featuring AI-powered IDEs like Cursor and enterprise coding agents.
• Reinforcement learning with verifiers (RLVR) is reshaping how model intelligence scales. Instead of just pumping huge volumes of data into bigger models, this new approach fine-tunes models with verifiable rewards — a perfect fit for coding where outputs can be objectively checked.
• Training models as “agents” capable of step-by-step reasoning and tool usage is transforming usefulness. Unlike traditional LLMs that provide single-shot answers, these agents can perform tasks interactively, integrating external tools like calculators and search engines. Anthropic led this charge with their model context protocol (MCP), greatly expanding functional capabilities and driving adoption.

Open-source models struggle to gain enterprise ground

While open-source LLMs like Meta‘s Llama remain popular, their share of enterprise AI workloads has actually declined slightly — from 19% to 13% in just six months. Despite launches by DeepSeek, Bytedance, and others, these models continue trailing the closed-source frontier by about nine to 12 months in performance.

There are advantages to open-source, including greater customization and on-prem deployment options. But the complexity in deploying these models and concerns around trust (especially for models from some Chinese companies) have slowed their uptake. Enterprises and startups alike are sticking with closed-source models to ensure top-tier performance.

Model upgrades beat switching: performance is king

Interestingly, switching between AI vendors is pretty rare nowadays. Instead, most enterprises and startups upgrade within their existing platforms to the newest model versions. For example, within a month of the Claude 4 release, 45% of Anthropic users migrated to the new model, while older versions rapidly lost share.

Performance is consistently prioritized over price or speed. Even as individual models drop sharply in cost, builders don’t use cheaper older models — they flock to the best-performing versions as soon as they’re available.

AI spending shifts gears: inference outpaces training

Another big trend is in how enterprises spend their AI compute budgets. There’s a clear shift from training models—which can be expensive and complex—to inference, where models are actually deployed and used in production.

Startups lead this trend, with 74% reporting that the majority of their compute usage is now for inference, up from 48% a year ago. Large enterprises are close behind, with nearly half of them saying most of their AI compute is dedicated to inference workloads.

What’s next for enterprise LLMs?

The pace of change in the AI market still feels dizzying, with new model breakthroughs, evolving economic models, and rapid shifts in what enterprises want driving constant flux. But it’s clear that we’re entering a phase ripe for building durable AI businesses on top of these foundational models.

Few things stand out to me from this mid-year update:

• Closed-source, high-performance models are winning enterprise trust and dollars. The gap between open vs. closed model performance and usability still matters a lot.
• Model capabilities are advancing along multiple dimensions, especially through agent architectures and reinforcement learning. This is expanding what AI can actually do.
• The economics of AI are shifting toward large-scale, inference-driven production use. This will likely influence infrastructure, tooling, and cost optimizations going forward.

As the landscape continues evolving, staying close to these trends is crucial — whether you’re building AI infrastructure, applications, or simply trying to navigate where value flows in the AI ecosystem.

Watching Anthropic’s ascent, the meaning of “agents,” and the ongoing tug-of-war between open and closed source has been genuinely eye-opening. It’s becoming clear that AI’s long game is not just about flashy breakthroughs — it’s about foundational shifts in how models are built, deployed, and monetized.

The post The rise of Anthropic and the shifting landscape of enterprise LLMs in 2025 appeared first on Aiholics: Your Source for AI News and Trends.

There’s a new kid on the block in the AI image creation space that’s causing quite a stir. Flux, Black Forest Labs‘ latest text-to-image models which build upon their famous Stable Diffusion engine.

Flux is available in three variants: Flux.1 [pro], Flux.1 [dev], and Flux.1 [schnell]. The best quality comes with the pro version since it is the flagship model. For non-commercial users, there is a dev version while schnell is the quickest; perfect for individuals who want to create images themselves.

But what makes flux so special? Well first of all, its enormous size does. It is one of the largest AI models out there with 12 billion parameters.

This size lets it comprehend and create images with extremely high detail levels and accuracy possible by any other AI up until now according to the team behind it, including such well-known ML models as Midjourney or DALL-E 3.

It’s amazing how well Flux listens! If you describe an image for it in words, this AI will do everything possible to accurately reproduce each word. Furthermore, it handles different sizes and shapes well as generating images with text also.

Black Forest isn’t just a newbie club; they are some of the most notable creators of recent AI imaging tools in this area today. They’ve got $31 million from investors like Andreessen Horowitz ready to take their groundbreaking creation to another level.

Nonetheless, flux represents only one small phase of their journey though. The team has already begun working on applying this technology to video production too where they promise top-notch editable videos that can be created fast just using texts alone.

To get started using flux you have some choices. The pro version may be accessed through an API whereas platforms like HuggingFace and GitHub offer access to dev or schnell versions respectively; therefore enabling both professionals and hobbyists to benefit from this thrilling new technology.

Test Flux in our site for free!

Making AI image creation more accessible, powerful, and imaginative than ever before is what Black Forest Labs intends to do with Flux. Tools like Flux are blazing the trail as AI continues to revolutionize our media creation and consumption practices, ushering us into a future where only imagination limits us.

The post Flux: The new powerhouse in AI image creation (Test it free in our website!) appeared first on Aiholics: Your Source for AI News and Trends.