Meta AI News - Page 12 of 17

Memory remains the silent choke point of large language models. As models balloon to hundreds of billions of parameters, the cost to serve them...

Meta AI’s latest brain-to-text model doesn’t read minds, it reads MEG signals, and it reads them disturbingly well.

AMD just matched Nvidia’s top chip. In a head-to-head sprint, the company's new MI355X accelerator fine-tuned a Llama 2 70B model in the same time as...

The old AI competition had a certain civility. That era is over. Google’s Gemini 3 Pro just landed a knockout blow in a street fight.

Meta pulled the plug on an internal leaderboard that tracked how much AI tokens its employees were burning through, after the company's AI spending...

Monday.com told the SEC on Wednesday that it's cutting about 20% of its staff, just over 600 people, folding the move into what it calls a...

Reproducibility in LLM orchestration has been a messy affair, hidden race conditions, hallucinated variables, and black-box behaviors that sabotage...

Quantizing a model? It's math, not magic. The real trick is running that math on Windows with a Radeon card.

Moonshot AI just scored a 71.3% on SWE-Bench Verified. That's not a minor bump. It beats OpenAI and Anthropic on a test of real-world coding.

NVIDIA just wired two billion dollars into the chip design business. The target is Synopsys, the company that makes the tools engineers use to design...

Meta released Muse Spark 1.2 this week, its latest open-weights model, paired with the company's first dedicated coding agent.

Singapore-based PixVerse closed an extended Series C round worth $439 million, pushing its valuation past $2 billion.

Meta burns through trillions of tokens a week on Microsoft's Azure cloud, and according to a Bloomberg report, that habit costs the company hundreds...

Yann LeCun has a word for the AI establishment: LLM-pilled. Since leaving Meta in November, the pioneering researcher has waged a public war on the...

Google wants its AI to be the thing you forget about. That’s the quiet pitch behind its new push into smart glasses.

India has ordered its internet to perform a magic trick. The government's new IT rules demand that platforms like Instagram and X block deepfakes and...

A.R. Rahman’s Secret Mountain isn’t just a band, it’s a living, breathing digital experiment. Now Google Cloud is supplying the neural backbone.

Most speech AI works for a handful of languages. Meta claims its new model now understands 1,600. The number is more believable than usual.

Two hundred dollars a month. That’s the price of a decent car payment, or a generous grocery run.

Local language models don’t need a cloud connection to do real work. I built a minimal Python agent running Llama 3.2 through Ollama, no external...