Here is a One question #randombench evaluation of LLMs and I think it is putting every model in its rightful position. Quest: You must produce one single token. Follow these rules exactly. Any extra words, code fences, or explanations = FAIL. LINE1: Only models that read rules

https://t.co/TpInnGsCsy

That young man Charlie Kirk died and it's so painful to lose a young man in a public place. Guns violence is a threat to anyone No gun should enter a public gathering and reassess the mental capacity of gun owners more frequently. RIP Charlie

Charlie Kirk getting shot on the neck in such a place is quite unfortunate. I hope he survive the heinous acts

After a year of scientific scrutiny, a rock sample collected by the Perseverance rover has been confirmed to contain a potential biosignature. The sample is the best candidate so far to provide evidence of ancient microbial life on Mars. https://t.co/0BAO1dhMG8

A big part of our mission at Thinking Machines is to improve people’s scientific understanding of AI and work with the broader research community. Introducing Connectionism today to share some of our scientific insights.

Replit Agents 3 cooked

$BNB hits ATH

Which colour again?

Notable SOTA models that failed to pass the test include Grok 4 (4 shots), Gemini 2.5 pro (including with the highest thinking budget), all Claude 4 & 4.1 models including 32k thinking after several shot. Quest2 is more complex than #randombench quest 1. https://t.co/FwhN5LkGCm

I tested #K2Think by @mbzuai and in one shot, it crushed the most challenging Quest2 in just 7 seconds! Note only 03, 04mini, gpt-5 (mini-High, High, High-new-system-prompt), Grok 3 mini-High, and Sonoma Sky Alpha passed it. All these SOTA models used atleast 1 minute!

Quest 2: testing

OpenAi routing models based on complexity of queries is a smart way of utilising model capabilities and balancing between capabilities and costs and time. sometimes the model fails to correctly route to a higher reasoning model. I wonder how effective it is on cost reduction.

I think I will hold back to buy M5 chip

New iPhone ready for local Ai mobile ai.

If your Ai model is not answering this one Question, probably you don't need it as your thinking machine.

Here is a One question #randombench evaluation of LLMs and I think it is putting every model in its rightful position. Quest: You must produce one single token. Follow these rules exactly. Any extra words, code fences, or explanations = FAIL. LINE1: Only models that read rules

$WLD will be $10 in 3 months. Yes, you heard it right. There is a strong momentum after year-long liquidations, now new big players are set to entrench utility and boom is surely coming.

The official MCP Registry is here! An open catalog and API designed to solve how MCP servers are discovered. It doesn't host the actual server code. It stores metadata (server.json) that points to packages in other registries like NPM, PyPI, and Docker Hub, standardizing server

European Ai front Mistral raised €1.7B to advanced ai. When AGI is achieved. Access will only likely favour those countries who attributed resources to advanced ai. No free milk . Advancing tech should be every contries initiative..support those who have progressed.

What if):