MeMo's memory model lets teams upgrade their LLM without retraining it — and performance jumps 26%

Researchers' MeMo keeps AI memory separate from reasoning, so teams can upgrade their LLM without retraining it and see a 26% ...

Timing Trick Cuts Energy Used in LLM Training by Up to 14 Percent

GPT-4, took an estimated 50 gigawatt-hours to train, or the equivalent of 5,000 American homes’ yearly power consumption.

Xiaomi AI and LLMs: Every model, every feature, everything you need to know

Xiaomi has always been known for affordable smartphones and smart home gadgets. But over the last year and a half, the ...

Routing Strategies: How AI Teams Select the Right Language Model

AI teams have more language model options available to them than at any point before. As that catalog has expanded, so ...
Crypto Briefing

MIT’s MeMo framework boosts LLM performance by 26% without retraining

MIT's MeMo framework trains a compact memory model that boosts LLM performance by up to 26.73% without retraining, with major implications for crypto AI agents.

Researchers say they trained a foundation model from scratch for about $1,500

Sapient researchers trained a 1B reasoning model on just 40B tokens — scoring competitively with 2B-7B models at a fraction ...

Anthropic releases Claude Mythos 5 as Fable 5 with restrictions

Claude Mythos 5 is available for the NSA and selected partners. The released, restricted version is called Claude Fable 5.