#ai-models
#ai-models

OpenEvidence v. Pathway: The Legal Battle Over AI Reverse Engineering

Generative AI models can be reverse engineered, raising legal concerns about trade secrets and competitive practices.

15 hours ago

Artificial intelligence

Freepik releases an 'open' AI image generator trained on licensed data | TechCrunch

Freepik launched F Lite, an AI image model trained solely on commercially licensed images, addressing copyright concerns in generative AI.

Meta Unveils Llama 4 AI Series Featuring New Expert-Based Architecture

Meta launched Llama 4, its first AI model series utilizing a mixture of experts architecture for improved resource efficiency.

Meta releases Llama 4, a new crop of flagship AI models | TechCrunch

Meta has launched the Llama 4 collection of AI models, pushing advancements with new architectures and expanded capabilities.

Meta says its latest AI models answer more 'contentious' questions than the last version

Llama 4 addresses more contentious topics with improved balance and lower refusal rates than its predecessor.

fromThe Verge

Meta AI gets two new models as Meta releases Llama 4

Meta's Llama 4 models reportedly outperform competitors like OpenAI and Google across diverse benchmarks.

Meta introduces Llama 4 with two new models available now, and two more on the way

Meta has launched Llama 4 models: Scout for summarization and Maverick for general chat use, with Behemoth expected to surpass all competitors soon.

18 hours ago

Meta is making it easier to use Llama models for app development

Meta introduces the Llama API to empower developers to leverage Llama models with ease and better control.

fromTechRepublic

Meta Unveils Llama 4 AI Series Featuring New Expert-Based Architecture

Meta launched Llama 4, its first AI model series utilizing a mixture of experts architecture for improved resource efficiency.

Meta releases Llama 4, a new crop of flagship AI models | TechCrunch

Meta has launched the Llama 4 collection of AI models, pushing advancements with new architectures and expanded capabilities.

Meta says its latest AI models answer more 'contentious' questions than the last version

Llama 4 addresses more contentious topics with improved balance and lower refusal rates than its predecessor.

fromThe Verge

Meta AI gets two new models as Meta releases Llama 4

Meta's Llama 4 models reportedly outperform competitors like OpenAI and Google across diverse benchmarks.

Meta introduces Llama 4 with two new models available now, and two more on the way

Meta has launched Llama 4 models: Scout for summarization and Maverick for general chat use, with Behemoth expected to surpass all competitors soon.

18 hours ago

Meta is making it easier to use Llama models for app development

Meta introduces the Llama API to empower developers to leverage Llama models with ease and better control.

more#meta

fromComputerworld

1 day ago

China's Alibaba and Baidu step up global competition with new reasoning-focused AI models

Chinese AI models are matching Western performance while being significantly cheaper, pressuring US firms to innovate and reduce costs.

#machine-learning

"It's a lemon"-OpenAI's largest AI model ever arrives to mixed reviews

GPT-4.5 may be OpenAI's last traditional model, with a shift to simulated reasoning due to high costs and diminishing returns.

Google's Gemma 3 is an open source, single-GPU AI with a 128K context window

Gemma 3 is highlighted as a top single-accelerator model by Google, excelling in chat capabilities and efficiency.

The hottest AI models, what they do, and how to use them | TechCrunch

AI model proliferation complicates tracking advancements and real-world applicability.

TechCrunch provides a curated overview of advanced AI models to guide users.

The hottest AI models, what they do, and how to use them | TechCrunch

AI models are evolving rapidly, making it difficult for users to keep track.

Performance benchmarks may not accurately reflect real-world applications of AI.

TechCrunch provides an updated overview of advanced AI models released since 2024.

OpenAI's new o3 and o4-mini models are all about 'thinking with images'

OpenAI's new models, o3 and o4-mini, enhance reasoning capabilities and introduce full tool access for improved problem-solving.

The hottest AI models, what they do, and how to use them | TechCrunch

The rapid advancement of AI models requires understanding their practical applications rather than just technical metrics.

"It's a lemon"-OpenAI's largest AI model ever arrives to mixed reviews

GPT-4.5 may be OpenAI's last traditional model, with a shift to simulated reasoning due to high costs and diminishing returns.

Google's Gemma 3 is an open source, single-GPU AI with a 128K context window

Gemma 3 is highlighted as a top single-accelerator model by Google, excelling in chat capabilities and efficiency.

The hottest AI models, what they do, and how to use them | TechCrunch

AI model proliferation complicates tracking advancements and real-world applicability.

TechCrunch provides a curated overview of advanced AI models to guide users.

The hottest AI models, what they do, and how to use them | TechCrunch

AI models are evolving rapidly, making it difficult for users to keep track.

Performance benchmarks may not accurately reflect real-world applications of AI.

TechCrunch provides an updated overview of advanced AI models released since 2024.

OpenAI's new o3 and o4-mini models are all about 'thinking with images'

OpenAI's new models, o3 and o4-mini, enhance reasoning capabilities and introduce full tool access for improved problem-solving.

The hottest AI models, what they do, and how to use them | TechCrunch

The rapid advancement of AI models requires understanding their practical applications rather than just technical metrics.

more#machine-learning

New AI models are running on just one or two chips. Will they trigger another DeepSeek moment for Nvidia?

New AI models require fewer Nvidia chips for increased performance.

Gemini "coming together in really awesome ways," Googler says after 2.5 Pro release

Google is catching up in generative AI with the Gemini 2.5 Pro release, aiming to compete with ChatGPT.

Google is shipping Gemini models faster than its AI safety reports | TechCrunch

Google has accelerated AI model launches to remain competitive in the evolving AI industry.

Concerns arise over Google's transparency and safety reporting for new models like Gemini 2.5 Pro.

The lack of model cards for recent releases contrasts with standards set by competitors.

Gemini 2.5 Pro is Google's most expensive AI model yet | TechCrunch

Gemini 2.5 Pro provides high performance in coding and reasoning but comes with elevated pricing, reflecting demand and computing costs in the AI space.

fromSearch Engine Roundtable

1 day ago

Google Used Search Signals & Data To Train Gemini AI Models

Google leverages search engine data to enhance AI models like Gemini, aiming to improve quality and accuracy in AI-generated responses.

4 weeks ago

Google's new experimental Gemini 2.5 model rolls out to free users

Gemini 2.5 Pro is now available for free users with some advanced features.

Gadgets

New AI models are running on just one or two chips. Will they trigger another DeepSeek moment for Nvidia?

New AI models require fewer Nvidia chips for increased performance.

Gemini "coming together in really awesome ways," Googler says after 2.5 Pro release

Google is catching up in generative AI with the Gemini 2.5 Pro release, aiming to compete with ChatGPT.

Google is shipping Gemini models faster than its AI safety reports | TechCrunch

Google has accelerated AI model launches to remain competitive in the evolving AI industry.

Concerns arise over Google's transparency and safety reporting for new models like Gemini 2.5 Pro.

The lack of model cards for recent releases contrasts with standards set by competitors.

Gemini 2.5 Pro is Google's most expensive AI model yet | TechCrunch

Gemini 2.5 Pro provides high performance in coding and reasoning but comes with elevated pricing, reflecting demand and computing costs in the AI space.

fromSearch Engine Roundtable

1 day ago

Google Used Search Signals & Data To Train Gemini AI Models

Google leverages search engine data to enhance AI models like Gemini, aiming to improve quality and accuracy in AI-generated responses.

4 weeks ago

Google's new experimental Gemini 2.5 model rolls out to free users

Gemini 2.5 Pro is now available for free users with some advanced features.

Why Tax Breaks for Data Centers Could Backfire on States

Tax breaks for data centers raise questions about their true economic impact and cost to states.

6 days ago

Lattica enters market with FHE for AI privacy issues

Lattica launches a cloud-based platform utilizing Fully Homomorphic Encryption for secure AI model usage in sensitive sectors.

#benchmarking

Crowdsourced AI benchmarks have serious flaws, some experts say | TechCrunch

Crowdsourced benchmarking platforms like Chatbot Arena face ethical criticism from experts regarding their effectiveness and validity in evaluating AI models.

Artificial intelligence

DeepSeek's V3 AI model gets a major upgrade - here's what's new

DeepSeek's new V3-0324 model shows significant improvements in reasoning and web development but is recommended for simpler tasks.

The AI startup aims to tackle benchmark saturation with advanced assessments.

Artificial intelligence

Too Many AIs With Too Many Terrible Names: How to Choose Your AI Model | HackerNoon

The AI model landscape is cluttered and confusing, leading to limited user engagement despite numerous advancements.

Crowdsourced AI benchmarks have serious flaws, some experts say | TechCrunch

Crowdsourced benchmarking platforms like Chatbot Arena face ethical criticism from experts regarding their effectiveness and validity in evaluating AI models.

Artificial intelligence

DeepSeek's V3 AI model gets a major upgrade - here's what's new

DeepSeek's new V3-0324 model shows significant improvements in reasoning and web development but is recommended for simpler tasks.

The AI startup aims to tackle benchmark saturation with advanced assessments.

Artificial intelligence

Too Many AIs With Too Many Terrible Names: How to Choose Your AI Model | HackerNoon

The AI model landscape is cluttered and confusing, leading to limited user engagement despite numerous advancements.

more#benchmarking

1 year ago

The Best AI Models For Invoice Processing: Benchmark Comparisons | HackerNoon

One AI model significantly outperformed others by 20% in processing invoices without fine-tuning.

#openai

Microsoft trains own AI models as alternative to OpenAI

Microsoft is seeking to lessen its reliance on OpenAI by developing its own AI models and exploring alternatives like xAI, Meta, and DeepSeek.

OpenAI finally unveils GPT-4.5. Here's what it can do

OpenAI's GPT-4.5 offers enhanced contextual understanding and emotional intelligence, marking a significant improvement over its predecessors.

AI firms follow DeepSeek's lead, create cheaper models with "distillation"

Distillation enables affordable access to AI models, benefiting developers and businesses using smaller devices like laptops and smartphones.

OpenAI to launch AI models that can think up their own experiments, says report

OpenAI is set to release new GPT-4.1 models capable of designing scientific experiments, focusing on inter-disciplinary knowledge synthesis.

fromTechRepublic

Have Some Spare Cash? You'll Need it for OpenAI's New API

OpenAI's o1-pro is a powerful, premium reasoning model for generative AI, available at high prices to cater to advanced users.

Access to future AI models in OpenAI's API may require a verified ID | TechCrunch

OpenAI will require organizations to complete an ID verification process to access advanced AI models.

Microsoft trains own AI models as alternative to OpenAI

Microsoft is seeking to lessen its reliance on OpenAI by developing its own AI models and exploring alternatives like xAI, Meta, and DeepSeek.

OpenAI finally unveils GPT-4.5. Here's what it can do

OpenAI's GPT-4.5 offers enhanced contextual understanding and emotional intelligence, marking a significant improvement over its predecessors.

AI firms follow DeepSeek's lead, create cheaper models with "distillation"

Distillation enables affordable access to AI models, benefiting developers and businesses using smaller devices like laptops and smartphones.

OpenAI to launch AI models that can think up their own experiments, says report

OpenAI is set to release new GPT-4.1 models capable of designing scientific experiments, focusing on inter-disciplinary knowledge synthesis.

fromTechRepublic

Have Some Spare Cash? You'll Need it for OpenAI's New API

OpenAI's o1-pro is a powerful, premium reasoning model for generative AI, available at high prices to cater to advanced users.

Access to future AI models in OpenAI's API may require a verified ID | TechCrunch

OpenAI will require organizations to complete an ID verification process to access advanced AI models.

more#openai

Google Gemini 2.5 Flash can think for as long as devs want it to

Google's Gemini 2.5 Flash offers a lighter alternative to previous models, enhancing user experience amid heightened competition in AI.

fromChristianheilmann

The rise of Model Fatigue - or is it just me?

AI news is often filled with hyperbole, lacking practical insights into the applicability of new models.

There is a need for more case studies instead of constant competition over model size and performance.

Practical applications and user experiences can better inform and engage people interested in AI.

#cost-efficiency

Amazon is working on a new 'reasoning' AI model that competes with OpenAI and Anthropic

Amazon is launching a new AI model with advanced hybrid reasoning capabilities by June.

The Nova model focuses on cost efficiency and aims to rank among the top 5 in performance benchmarks.

Artificial intelligence

GPT-4.1's 1M-token Context Window is Impressive but Insufficient for Real-world Use Cases | HackerNoon

Large context windows in AI are attractive but come with hidden costs and complexities that impact their practical usability.

Amazon is working on a new 'reasoning' AI model that competes with OpenAI and Anthropic

Amazon is launching a new AI model with advanced hybrid reasoning capabilities by June.

The Nova model focuses on cost efficiency and aims to rank among the top 5 in performance benchmarks.

Artificial intelligence

GPT-4.1's 1M-token Context Window is Impressive but Insufficient for Real-world Use Cases | HackerNoon

Large context windows in AI are attractive but come with hidden costs and complexities that impact their practical usability.

more#cost-efficiency

fromtechcrunch.com

Artificial intelligence

Debates over AI benchmarking have reached Pokemon

AI benchmarks, including Pokemon, are complicated by implementation differences and custom tools that influence performance outcomes.

#artificial-intelligence

The future of AI isn't the model-it's the system

The model alone is insufficient; practical value comes from its surrounding systems and capabilities.

fromFinbold

Orb AI: The Search Engine of The Future Has Arrived - Presale Raises $1 Million In 3 Weeks

Orb AI connects top AI models to revolutionize digital content generation.

The most innovative companies in artificial intelligence for 2025

The AI industry is transitioning towards real-time reasoning models as a crucial step towards achieving artificial general intelligence.

Artificial intelligence

DeepSeek: Everything you need to know about the AI chatbot app | TechCrunch

DeepSeek's AI chatbot became a top app, raising concerns about U.S. dominance in AI.

The future of AI isn't the model-it's the system

The model alone is insufficient; practical value comes from its surrounding systems and capabilities.

fromFinbold

Orb AI: The Search Engine of The Future Has Arrived - Presale Raises $1 Million In 3 Weeks

Orb AI connects top AI models to revolutionize digital content generation.

The most innovative companies in artificial intelligence for 2025

The AI industry is transitioning towards real-time reasoning models as a crucial step towards achieving artificial general intelligence.

more#artificial-intelligence

Artificial intelligence

DeepSeek: Everything you need to know about the AI chatbot app | TechCrunch

DeepSeek's AI chatbot became a top app, raising concerns about U.S. dominance in AI.

#open-source

'Open' model licenses often carry concerning restrictions | TechCrunch

Gemma 3's licensing poses significant legal risks for commercial use, raising concerns among developers about long-term viability.

China puts American AI industry on notice yet again with 'Ernie X1,' Baidu's new open-source reasoning model

Baidu's new AI models aim to rival OpenAI and DeepSeek in performance while being significantly cheaper.

fromHubspot

3 months ago

How Qwen 2.5 Just Beat the Top AI Models - HubSpot SVP of Marketing Shares The Industry Impact

Qwen 2.5 is an open-source, multimodal AI model that outperforms others in complex problem-solving and is available for free.

'Open' model licenses often carry concerning restrictions | TechCrunch

Gemma 3's licensing poses significant legal risks for commercial use, raising concerns among developers about long-term viability.

China puts American AI industry on notice yet again with 'Ernie X1,' Baidu's new open-source reasoning model

Baidu's new AI models aim to rival OpenAI and DeepSeek in performance while being significantly cheaper.

fromHubspot

3 months ago

How Qwen 2.5 Just Beat the Top AI Models - HubSpot SVP of Marketing Shares The Industry Impact

Qwen 2.5 is an open-source, multimodal AI model that outperforms others in complex problem-solving and is available for free.

more#open-source

#technology

fromITPro

We need to stop caring about AI model releases

Focus on the practical benefits of AI models rather than just technical specifications.

The excitement around model launches should center on their usefulness for users, not their technical details.

fromTearsheet

London startup

The Quarterly Review: Current's CTO Trevor Marshall reports on Model-Driven Success - Tearsheet

Current is enhancing model-driven deployments to improve consumer product offerings and performance.

The firm has advanced its AI systems to improve transaction risk detection and combat fraud.

fromITPro

We need to stop caring about AI model releases

Focus on the practical benefits of AI models rather than just technical specifications.

The excitement around model launches should center on their usefulness for users, not their technical details.

fromTearsheet

London startup

The Quarterly Review: Current's CTO Trevor Marshall reports on Model-Driven Success - Tearsheet

Current is enhancing model-driven deployments to improve consumer product offerings and performance.

The firm has advanced its AI systems to improve transaction risk detection and combat fraud.

more#technology

A new, challenging AGI test stumps most AI models | TechCrunch

The new ARC-AGI-2 test challenges AI models with puzzle-like problems to measure their general intelligence more effectively than previous tests.

Digital life

fromWIRED

Synchron's Brain-Computer Interface Now Has Nvidia's AI

Synchron's tech allows non-invasive brain-computer interface development, enabling data collection for AI models using avatars as time stamps.

fromInvestor's Business Daily

Meta Stock Slides To Approach 3-Month Low Despite This AI Milestone

Meta stock declines despite AI milestone of 1 billion downloads for Llama models.

Concerns over economic conditions impact investor sentiment on tech stocks, including Meta.

#deepseek

US considers banning DeepSeek on government devices

DeepSeek's AI models raise user privacy concerns as they gain popularity, particularly with the R1 model outpacing OpenAI's algorithm.

Artificial intelligence

Baidu launches two new versions of its AI model Ernie | TechCrunch

Baidu launches new AI models Ernie 4.5 and Ernie X1 with advanced capabilities and cost-effective positioning.

US considers banning DeepSeek on government devices

DeepSeek's AI models raise user privacy concerns as they gain popularity, particularly with the R1 model outpacing OpenAI's algorithm.

fromCreativeApplications.Net

Artificial intelligence

Baidu launches two new versions of its AI model Ernie | TechCrunch

Baidu launches new AI models Ernie 4.5 and Ernie X1 with advanced capabilities and cost-effective positioning.

more#deepseek

Online Community Development

Habsburg AI Portrait Studies

Habsburg AI Portrait Studies examines how generative portraiture amplifies aesthetic characteristics of AI models, revealing structural implications of synthetic data usage.

ChatGPT doubled its weekly active users in under 6 months, thanks to new releases | TechCrunch

ChatGPT experienced accelerated growth, reaching 400 million weekly active users partly due to new models and features.

European startups

Google DeepMind, Cohere, and Twelve Labs join Sessions: AI | TechCrunch

AI model development is rapidly evolving, presenting unique opportunities for startup founders to build and innovate in the field.

Open AI, Anthropic invite US scientists to experiment with frontier models

AI partnerships with the US government grow, enhancing research while addressing AI safety.

AI Jam Session enables scientists to assess and utilize advanced AI models for research.

NYC startup

fromtechcrunch.com

Anthropic raises $3.5B to fuel its AI ambitions

Anthropic raised $3.5 billion to enhance AI systems and user experience while aiming for substantial growth despite high development costs.

fromMedium

The business challenge for AI-native applications

AI integration is crucial for building industry-defining products.

The current AI competitive landscape includes AI-native applications and model providers.

5 years ago

What Is a Diffusion LLM and Why Does It Matter? | HackerNoon

Inception Labs launched Mercury Coder, the first commercial diffusion LLM, promising faster processing speeds and innovative capabilities.

fromInfoWorld

What is retrieval-augmented generation? More accurate and reliable LLMs

RAG enhances the accuracy of large language models by integrating external data sources, but it isn't a comprehensive solution.

fromInfoWorld