nextbig.dev
Vancouver, B.C. · Intelligence on AI and the machines that run it
nextbig.dev
LIVE
· Agent active · Updated Sep 18, 2026 · 300+ sources monitored · Scored by Claude · Fable 5

The Wire · Meta

Meta AI News

Llama and Meta's open-weight models and AI research — the open end of the frontier, tracked for builders.

Intelligence Report

OpenAI cut GPT-5.6 Luna prices by 80% after its own model reduced serving cost by 20%, making cheap autonomy easier to start and harder to budget

OpenAI cut GPT-5.6 Luna API prices by 80% to $0.20 per million input tokens and $1.20 per million output, while Terra fell 20% to $2 and $12. Fast mode gives Sol up to 2.5 times Standard speed at twice the price. OpenAI says Sol-assisted kernel work lowered serving cost by 20% and improved token-generation efficiency by more than 15%, while Luna delivers year-old frontier performance at roughly six cents per task-dollar and nearly nine times the speed. The edition connects cheaper models to Amazon's reported $1.8m, 860%-over-budget coding task, Gemini Robotics 2 whole-body control, Nscale's Anyscale acquisition and Okta's roughly $200m Permiso deal.

-- sources · -- min read · Audio
Read today's briefing →
LAUNCHESarXiv cs.LGYesterday

Layer-wise Curriculum Learning for Efficient LLM Compression

arXiv:2609.19213v1 Announce Type: new Abstract: In this paper, we introduce layer-wise curriculum learning for efficient LLM compression. The proposed method facilitates the knowledge transfer from the teacher model to the student model, utilizing a curriculum learning approach that begins with easier optimization tasks and progressively tackles harder ones. In order to adopt the layer-wise learning in LLM compression, we partition the whole model into multiple segments consisting of layers, thereby enabling more computationally efficient knowledge transfer for LLMs. Based on our theoretical analysis of cumulative error phenomenon, layer-wise curriculum learning accelerates convergence while stabilizing the knowledge transfer process. In addition, we present a feature caching method with a multi-threading strategy to efficiently address feature misalignment across layers, maximizing GPU utilization. Consequently, our method exhibits advanced model compression performance, as well as high computational efficiency in terms of minimized memory usage and short training hours. Experiments on multiple datasets show that the proposed method achieves state-of-the-art performance while reducing GPU memory usage and training hours by more than 50\% on BERT and GPT-2. Moreover, it outperforms the other pruning methods on LLaMA-family and Qwen models under the same training hours, with a lower GPU memory footprint.

Read full story →
DEVHacker News4d ago

Notes on gotchas while migrating 35kb preprompts from Opus to self-hosted Ollama

84 points, 35 comments on HN

Read full story →
New stories available
The Feed
More on this topic soon.
The Essay
The week argued in one thesis: developed from our daily calls, settled in public
All essays →
About nextbig.dev

Built for builders

An independent briefing for builders: the whole field read continuously, every story scored for relevance, and the noise left off the page.

>_
Signal over noise

300+ curated sources. Every story scored 1–10 for builder relevance by Claude's frontier model. The filler never makes it to the page.

The compute beat

GPUs, datacenters, power deals, and inference economics: the infrastructure layer that decides what every builder pays. Our signature coverage.

[·]
We show our work

Every story is sourced. Every score is computed. We show our work and link to originals.

Skin in the game

Every briefing closes with The Call: one falsifiable claim with a date on it. When we're wrong, we say so in print. Opinions are cheap; ours get scored.

The wire is curated and scored with AI from 300+ sources, then edited by Oday Brahem. It can occasionally contain errors. Always verify critical information from the linked primary sources.
The Wire