NextBigFuture

OpenAI Strawberry LLM Reasoning Needs More Compute and Energy for Inference

Jim Fan is one of Nvidia’s senior AI researchers. The shift could be about many orders of magnitude more compute and energy needed for inference that can handle the improved reasoning in the OpenAI ...
HHS

How Test-Time Compute Can Help Scale AI

For years, it seemed obvious that the best way to scale up artificial intelligence models was to throw more upfront computing resources at them. The theory was that performance improvements are ...
VentureBeat

How test-time scaling unlocks hidden reasoning abilities in small language models (and allows them to outperform LLMs)

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now Very small language models (SLMs) can ...
EDN

Purpose-built AI inference architecture: Reengineering compute design

Over the past several years, the lion’s share of artificial intelligence (AI) investment has poured into training infrastructure—massive clusters designed to crunch through oceans of data, where speed ...
Geeky Gadgets

Google DeepMind Unlocks the Future of AI Efficiency

Google DeepMind’s recent research offers a fresh perspective on optimizing large language models (LLMs) like OpenAI’s ChatGPT-o1. Instead of merely increasing model parameters, the study emphasizes ...
Business Insider

Google DeepMind researchers think they found a solution to AI's 'peak data' problem

You're currently following this author! Want to unfollow? Unsubscribe via the link in your email. Follow Alistair Barr Every time Alistair publishes a story, you’ll get an alert straight to your inbox ...
NextBigFuture

Cogito v2 – Inference-time search and New AI Self-improvement

The largest Cogito v2 671B MoE model is amongst the strongest open models in the world. It matches/exceeds the performance of the latest DeepSeek v3 and DeepSeek R1 models both, and approaches closed ...
The Next Platform

The Battle Begins For AI Inference Compute In The Datacenter

The major cloud builders and their hyperscaler brethren – in many cases, one company acts like both a cloud and a hyperscaler – have made their technology choices when it comes to deploying AI ...
EDN

Analog in-memory compute tackles the AI inference conundrum

An analog in-memory compute chip claims to solve the power/performance conundrum facing artificial intelligence (AI) inference applications by facilitating energy efficiency and cost reductions ...
Morningstar

Silicon Valley Gen AI Inference Compute Provider Tensordyne Achieves IDCA G2 Certification

SUNNYVALE, Calif., Sept. 19, 2025 /PRNewswire/ -- International Data Center Authority (IDCA) reports that Tensordyne has been able to achieve the G2 certification level after successfully completing ...
VentureBeat

When AI reasoning goes wrong: Microsoft Research shows more tokens can mean more problems

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now Large language models (LLMs) are ...