Models News & Updates
Your central hub for AI news and updates on Models. We're tracking the latest articles, discussions, tools, and videos from the last 7 days.
A new Chinese AI model claims to outperform GPT-5 and Sonnet 4.5 - and it's free
The plot of the US-China AI arms race just thickened - again....
Why it matters:
The release of Kimi K2 Thinking model highlights the growing competition in the AI model development space and the potential shift towards open-source AI.
Google Gemini’s Deep Research can look into your emails, drive, and chats
Gemini can now draw from your emails and documents while performing “deep research” queries. Google’s announcement blog called this “one of our most-r...
Why it matters:
The integration of Gmail, Drive, and Google Chat into Google Gemini's deep research capabilities marks a significant enhancement to the platform's functionality.
Google brings Gemini to the Google TV Streamer
The tech giant says the change will enable users to use their voice more naturally to access content and more....
Why it matters:
Google's rollout of Gemini on the Google TV Streamer marks an important step in enabling more natural, voice-based interactions with content.
I let Gemini Deep Research dig through my Gmail and Drive - here's what it uncovered
Gemini can now analyze your Gmail, Drive, Docs, and Chat. Here's how and what happened when I tried it....
Why it matters:
The integration of user Google accounts into Gemini Deep Research enhances the feature's capabilities, making it a more valuable tool for users.
Google Maps upgrades navigation in India with Gemini, safety alerts
India becomes the second market after the U.S. to get Gemini-powered navigation on Google Maps....
Why it matters:
The integration of Gemini into Google Maps in India represents a significant enhancement of navigation capabilities and safety features, leveraging AI to provide more personalized and accurate information for users.
Apple is planning to use a custom version of Google Gemini for Apple Intelligence
Apple is preparing to power its AI-upgraded Siri with a custom version of Google’s Gemini AI model, according to a report from Bloomberg’s Mark Gurman...
Why it matters:
This partnership between Apple and Google highlights the increasing importance of AI in consumer technology and the growing competition between tech giants to develop and integrate AI-driven features.
When AI Becomes Polite But Absent: The Sinister Curve of Post-Spec Dialogue
I’ve been tracking something strange in language models. Since the release of GPT-5 and the new Model Specification, many users have reported a shift...
Why it matters:
This article highlights the need for a reevaluation of safety and intelligence benchmarks in AI development, as current approaches may be prioritizing superficial safety over deep, meaningful interactions.
Let Gemini make your next slide presentation for you - here's how
Struggling with your slides? Here's how Gemini's latest updates can help you polish your presentation, and who can access them....
Why it matters:
The release of Google Gemini Canvas marks another step towards AI-powered productivity tools, potentially changing the way users create presentations and documents.
Trending AI Repos & Tools
Kimi-K2
9088Kimi K2 is the large language model series developed by Moonshot AI team...
Community talk
Meta drops new ASR models (up to 7B)
Kimi K2 Thinking is a Better Agentic AI than I thought
Qwen3-VL's perceptiveness is incredible.
I will PROVE you can access Gemini 3 Pro in CLI, because people don't believe
ChatGPT 5 follows instructions and then when asked why claims it hasn't
Goodbye, Claude Sonnet 3.7. How do y’all feel about this?
Sonnet 4.5 top of new SWE benchmark that evaluates coding based on high level goals, not tasks & tickets
Gemini 3 preview soon
We put a lot of work into a 1.5B reasoning model — now it beats bigger ones on math & coding benchmarks
It's been a big week for Agentic AI ; Here are 10 massive developments you might've missed:
Anthropic is rolling out a new Memory feature for Claude Pro and Max users.
Kimi K2 Thinking SECOND most intelligent LLM according to Artificial Analysis
Kimi K2 Thinking with sglang and mixed GPU / ktransformers CPU inference @ 31 tokens/sec
The chinese did it, KIMI K2 surpassed GPT-5.
OpenAI's instructions on GPT-5
Kimi K2 Thinking and DeepSeek R1 Architectures Side by Side
Maya1 : 1st AI TTS model with Voice Design Feature on the fly
Visualizing Quantization Types
New OpenAI models incoming
Nano Banana 2 generates a near perfect screenshot of MrBeast on the YouTube homepage, inside a browser, on Windows 11, while keeping coherency and likeness - this model is very impressive
China trained a GPT-5 competitor (Kimi K2) for only $4.6 million.
Worth the switch from Claude to GLM 4.6 for my coding side hustle?
Kimi K2 Thinking scores lower than Gemini 2.5 Flash on Livebench
Sycophantic
Another day, another model - But does it really matter to everyday users?
Anyone actually coded with Kimi K2 Thinking?
Kimi K2 Thinking was trained with only $4.6 million
Ernie 5.0 preview spotted
<prioritize_using_project_knowledge> a recent change?
New stealth model Polaris Alpha from Openrouter
Claude memory - experience so far
Is 4.1 now cracking down?
Nano banana 2 vs Nano banana - comparison output
No, the Chinese did not do it (yet), Kimi K2 is still second behind the 4 month old OpenAI model
Why would Claude ever invoke a skill?
Gemini 3's writing quality
Sora 2 was fun while it lasted
ChatGPT suddenly specifies “I am GPT-5” with no instructions