Le Chonk pairs 49 billion active parameters with a mixture of experts design and solves 93% of challenges on the Cybench security benchmark, Mistral said, with full open weights promised by the end of October.
Mistral released Large 4 on Oct. 6, a 1 trillion parameter model that uses a mixture of experts design with 49 billion active parameters, the company said in its announcement. The model, nicknamed Le Chonk, is available now in public preview through the Mistral Studio API, with full open weights due by the end of... Continue reading
The new endpoint returns probability scores, fixed choices or ordered ratings about 10 times faster than OpenAI's Responses API, aimed at classification and routing tasks, the company's developer docs say.
The release spans 372 families of results including progress tied to the Riemann hypothesis and a faster integer multiplication method, though OpenAI says some unformalized results could still contain errors.
The Nvidia-backed cloud company would be valued at $14.5 billion before the new money, with Coatue and Blackstone leading a round shaped partly by a $35 billion compute commitment from Anthropic.
The 740 million parameter model maps text, images, audio and video into one embedding space under an Apache 2.0 license, and Google says it needs as little as 191 megabytes of memory to run.
Monthly commits grew from 1.3 billion to 7.38 billion in a year and pushes nearly quintupled, pushing GitHub to move repository storage to Azure Blob Storage and separate reads from writes, the company said.
Common Sense Media, a nonprofit that offers reviews of apps, services, and entertainment with a focus on youth safety, today said that OpenAI's ChatGPT for Teens is an "unacceptable risk." ChatGPT for Teens, introduced...
Instead of helping marketers manage and optimize ad spend, the company is focusing on building the tools that generate the creative assets and campaigns.
Flai's revenue has grown 20x in a year, and the company just closed a $27 million Series A funding round.
There isn’t a single manual to read or prompt to give an LLM that can equip you with the skills and knowledge to build a company. But on November 4, TechCrunch Founder Summit gives founders the next-best thing: A packed...
Called Explore, Glimpse's new image processing software lets hardware makers speed up quality control checks using CT scanners.
Type One Energy is betting that its lean approach to fusion power will get a power plant on the grid faster, and investors have rewarded it with $200 million.
Just a couple of months after its last big raise, the AI chip startup is already being plied with investment offers at double or more its current value, sources tell TechCrunch.
Personal AI agents promise to shop, book flights, and make reservations for you. But deliberate blocks and anti-bot defenses are getting in the way, leaving consumers caught in the middle. A new standard aims to help.
I was a graph theory junkie long ago and even moved to Budapest for awhile to study among the greats. While I was there I started working on Barnette's Conjecture which came to occupy my thoughts over the next 24 years...
Release: datasette-atom 0.11a0 A minor fix for compatibility with the latest Datasette alphas. This meant we could upgrade the datasette.io site to Datasette 1.0a41. Tags: atom , datasette
TIL: Using Parseable with Datasette for OpenTelemetry traces I saw Parseable in a Show HN today - it's a new observability platform with both an open source (AGPL) Rust implementation (a single ~180MB binary), an...
The storied Silicon Valley firm makes its first pure defense investment.
Flow policies have become a standard policy class for learning robot behaviors from demonstrations, but reinforcement learning is still critical for improving pre-trained flow policies or learning them from scratch...
Conformal prediction is a popular tool for uncertainty quantification that outputs prediction sets with finite-sample coverage guarantees. While prediction set size is commonly used as a heuristic measure of...
Existing methods for 4D hand-object reconstruction often rely on costly per-sequence optimization, while generative approaches typically synthesize interactions from random noise, which can lead to unstable interaction...
Scientific research often begins by synthesizing ideas from a set of related papers to identify gaps and formulate new directions. However, training language models to perform this form of literature-grounded ideation...
Large language models (LLMs) can solve many narrow tasks, but querying them separately for millions of related instances can be prohibitively expensive. Can LLM agents autonomously create cheaper solutions for such...