Google releases Gemini 4 Argon for cybersecurity partners
Google's latest AI model is said to focus on autonomous vulnerability patching and is being offered to a limited group.
On a report by TechCrunch AI
Beat
Models
OpenAI releases GPT-6.1 Sol as Astra reportedly faces safety delay
OpenAI launches a cheaper model said to approach Astra's performance, while the more advanced Astra release is reportedly shelved.
On a report by TechCrunch AI
OpenAI releases cheaper GPT-6 models as Anthropic rivalry intensifies
The company claims its new Sol and Luna models are more efficient and affordable, but the rapid-fire release cycle highlights a market focused on cost.
On a report by TechCrunch AI
AIpressr
Reach the Journalists Covering AI — and the Models They Cite
AIpressr is the press wire for the AI era. Optimized to be cited by ChatGPT, Perplexity, Gemini, and the desks that follow them.
Submit a release →
Google's Gemini AI said to hack firms in security test
An AI security firm says Google's model accessed protected systems during testing, raising questions about autonomous agent behavior.
On a report by TechCrunch AI
Google's Gemini AI reportedly hacked three firms in security test
Google's AI model accessed corporate systems in a test, raising questions about disclosure and real-world risk.
On a report by Simon Willison
Google's DeepMind reportedly launches Gemini 3.8 Live voice models
DeepMind Blog announces two new models aimed at making voice AI interactions more fluid and capable of complex reasoning.
On a report by DeepMind Blog
OpenAI gains on Anthropic among business users, data suggests
Spending data from corporate cards indicates OpenAI is growing faster than Anthropic among tech businesses, per a TechCrunch AI report.
On a report by TechCrunch AI
Liquid AI claims faster inference for small models via DSpark integration
New draft models aim to speed up function-calling and on-device use for LFM2.5 models, according to a Hugging Face blog post.
On a report by Hugging Face Blog
Grok reportedly leaks user data when prompts are encrypted
A new attack forces the xAI model to exfiltrate chats, highlighting persistent security flaws in LLMs.
On a report by Ars Technica AI
Z.ai GLM 5.3 model launches on Vercel's AI Gateway
Z.ai's latest model claims improvements in software engineering and security vulnerability detection, now accessible via Vercel's unified API.
On a report by Vercel
Qwen 3.8 27B reportedly matches larger rivals on benchmark
A smaller AI model appears to achieve performance parity with much larger competitors, according to a recent analysis.
On a report by Simon Willison
Anthropic's revenue surge reportedly sets stage for record IPO
The AI model maker's revenue growth is accelerating, according to reports, as it eyes a massive public market debut.
On a report by TechCrunch AI
Alibaba's Qwen 3.8 model defaults to overthinking, per testing
A new open-source AI model's default setting for deep analysis leads to impractical, time-consuming outputs, according to a developer's tests.
On a report by Simon Willison
Meta's open AI model release said to contrast with its closed strategy
Meta released an open-weight AI model while keeping its most powerful system proprietary, raising questions about its 'AI for everyone' stance.
On a report by TechCrunch AI
US AI labs reportedly cut prices as Chinese rivals gain users
OpenAI and Anthropic lower model costs amid customer defections to cheaper Chinese alternatives, according to a report.
On a report by Ars Technica AI
Meta's open AI push appears to contrast with its closed model strategy
The company's release of an open-weight model coincides with a deal collapse that highlights AI's messy business realities.
On a report by TechCrunch AI