
Just as there are widely understood empirical laws of nature - for example, what goes up must come down, or every action has an equal and opposite reaction - the field of AI was long defined by a single idea: that more compute, more training data and more parameters makes a better AI model.
However, AI has since grown to need three distinct laws that describe how applying compute resources in different ways impacts model performance. Together, these AI scaling laws - pretraining scaling, post-training scaling and test-time scaling, also called long thinking - reflect how the field has evolved with techniques to use additional compute in a wide variety of increasingly complex AI use cases.
The recent rise of test-time scaling - applying more compute at inference time to improve accuracy - has enabled AI reasoning models, a new class of large language models (LLMs) that perform multiple inference passes to work through complex problems, while describing the steps required to solve a task. Test-time scaling requires intensive amounts of computational resources to support AI reasoning, which will drive further demand for accelerated computing.
What Is Pretraining Scaling? Pretraining scaling is the original law of AI development. It demonstrated that by increasing training dataset size, model parameter count and computational resources, developers could expect predictable improvements in model intelligence and accuracy.
Each of these three elements - data, model size, compute - is interrelated. Per the pretraining scaling law, outlined in this research paper, when larger models are fed with more data, the overall performance of the models improves. To make this feasible, developers must scale up their compute - creating the need for powerful accelerated computing resources to run those larger training workloads.
This principle of pretraining scaling led to large models that achieved groundbreaking capabilities. It also spurred major innovations in model architecture, including the rise of billion- and trillion-parameter transformer models, mixture of experts models and new distributed training techniques - all demanding significant compute.
And the relevance of the pretraining scaling law continues - as humans continue to produce growing amounts of multimodal data, this trove of text, images, audio, video and sensor information will be used to train powerful future AI models.
Pretraining scaling is the foundational principle of AI development, linking the size of models, datasets and compute to AI gains. Mixture of experts, depicted above, is a popular model architecture for AI training. What Is Post-Training Scaling? Pretraining a large foundation model isn't for everyone - it takes significant investment, skilled experts and datasets. But once an organization pretrains and releases a model, they lower the barrier to AI adoption by enabling others to use their pretrained model as a foundation to adapt for their own applications.
This post-training process drives additional cumulative demand for accelerated computing across enterprises and the broader developer community. Popular open-source models can have hundreds or thousands of derivative models, trained across numerous domains.
Developing this ecosystem of derivative models for a variety of use cases could take around 30x more compute than pretraining the original foundation model.
Developing this ecosystem of derivative models for a variety of use cases could take around 30x more compute than pretraining the original foundation model.
Post-training techniques can further improve a model's specificity and relevance for an organization's desired use case. While pretraining is like sending an AI model to school to learn foundational skills, post-training enhances the model with skills applicable to its intended job. An LLM, for example, could be post-trained to tackle a task like sentiment analysis or translation - or understand the jargon of a specific domain, like healthcare or law.
The post-training scaling law posits that a pretrained model's performance can further improve - in computational efficiency, accuracy or domain specificity - using techniques including fine-tuning, pruning, quantization, distillation, reinforcement learning and synthetic data augmentation.
Fine-tuning uses additional training data to tailor an AI model for specific domains and applications. This can be done using an organization's internal datasets, or with pairs of sample model input and outputs.
Distillation requires a pair of AI models: a large, complex teacher model and a lightweight student model. In the most common distillation technique, called offline distillation, the student model learns to mimic the outputs of a pretrained teacher model.
Reinforcement learning, or RL, is a machine learning technique that uses a reward model to train an agent to make decisions that align with a specific use case. The agent aims to make decisions that maximize cumulative rewards over time as it interacts with an environment - for example, a chatbot LLM that is positively reinforced by thumbs up reactions from users. This technique is known as reinforcement learning from human feedback (RLHF). Another, newer technique, reinforcement learning from AI feedback (RLAIF), instead uses feedback from AI models to guide the learning process, streamlining post-training efforts.
Best-of-n sampling generates multiple outputs from a language model and selects the one with the highest reward score based on a reward model. It's often used to improve an AI's outputs without modifying model parameters, offering an alternative to fine-tuning with reinforcement learning.
Search methods explore a range of potential decision paths before selecting a final output. This post-training technique can iteratively improve the model's responses
Most recent headlines
05/01/2027
Worlds first 802.15.4ab-UWB chip verified by Calterah and Rohde & Schwarz to be ...
07/10/2026
Dalet, a leading technology and service provider for media-rich organizations, today announced the latest Long-Term Supported (LTS) release of Dalet Flex. Build...
06/09/2026
June 9 2026, 23:00 (PDT) Dolby and MagentaTV Bring Fans Closer to the FIFA Worl...
05/08/2026
RSNs, teams, leagues, and streamers explore the present and future of local spor...
05/08/2026
As sports-production workflows generate larger volumes of unstructured data and ...
05/08/2026
Three aspiring broadcasters will contribute on-camera and inside the production ...
05/08/2026
Adam Meeks attends the premiere of his film Union County, an official selection of the 2026 Sundance Film Festival. (Photo by Jemal Countess/Sundance Institut...
05/08/2026
Major discounts, updates & remakes of early VST plug-ins
Steinberg are currently celebrating two significant anniversaries, with this week marking the 30th ...
05/08/2026
Flagship monitor range gains wireless remote control
KRK's flagship studio monitor range has just been upgraded, with three new fifth-generation models ...
05/08/2026
Rohde & Schwarz launches next phase of Airways New Zealand Air Traffic Control s...
05/08/2026
Rohde & Schwarz presents the R&S MP007, a man-portable tactical direction findin...
05/08/2026
Advertisers invested $53.6 million in the retirement sector in the last 12 month...
05/08/2026
Nielsen CMI and Ad Intel data reveals who's visiting IKEA, where else they s...
05/08/2026
Media buyers and sellers can now compare deduplicated campaign reach across computer, mobile, connected TV and linear TV across all measurable CTV publishers in...
05/08/2026
Stranger Things' Leads All Streaming Titles with 23.3 Billion Minutes in Fir...
05/08/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
05/08/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
05/08/2026
QuickLink, the leading provider of award-winning video production and remote guest integration solutions, announces the appointment of Kane Peterson to its Nort...
05/08/2026
XR research reveals marketers believe they have control over production costs, yet budget overruns, campaign delays and creative cutbacks remain commonplace
Se...
05/08/2026
Game On Live Studio Uses Blackmagic Design to Stream Tropical 7s
Brie Clayton August 4, 2026
0 Comments
Every match of world's largest youth rugby...
05/08/2026
Chris Fitzgerald to Head TDC's Experiential Growth
Brie Clayton August 4, 2026
0 Comments
Technical Direction Company promotes Chris Fitzgerald to...
05/08/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
05/08/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
05/08/2026
Profuz Digital is set to launch a powerful upgrade to its Profuz LAPIS digital media management platform at IBC this year at Stand 8.A88. France Galop, Canal+ F...
05/08/2026
KRK, a Gibson brand and leader in professional studio monitoring for nearly four decades, launches its new V Series Five Studio Monitors, the latest evolution o...
05/08/2026
Network expansion, continued customer growth and stable financial performance characterized Magyar Telekoms business results for the first half and second quart...
05/08/2026
Wednesday 5 August 2026
The Sky Foundation launches new bursary programme to su...
05/08/2026
Charlie Sheen Launches Mostly Sheenius in Partnership with Red Seat Ventures After Decades as One of the World's Most Recognizable and Interviewed Celebr...
04/08/2026
Live production in the cloud is a growing trend across the industry, and Major L...
04/08/2026
LAFC co-managing owner will succeed Don Garber on Jan. 1, 2027, as league prepar...
04/08/2026
In 2022, when the PGA TOUR was breaking ground on its new, state-of-the-art broa...
04/08/2026
Provided by POVORA, the new camera angle offers a dealer's-eye view of the c...
04/08/2026
Spiideo will provide the camera setup, with NEP handling broadcast integration and off-site video operations...
04/08/2026
Production team streamed more than 750 youth rugby matches using ATEM switchers,...
04/08/2026
Agreement gives KTC early access to NDI's Advanced SDK for integration across commercial displays, digital signage, and interactive panels...
04/08/2026
Twelve projects across four categories recognize collaborative work in productio...
04/08/2026
Multi-year agreement will introduce SkeleTRACK, semi-automated offside, automate...
04/08/2026
Three mc 56 consoles, dual A__UHD Core engines, RAVENNA networking, and HOME management supported live sound, monitoring, and international broadcast production...
04/08/2026
New contribution products support ultra-low-latency video transport across fixed, remote, and mobile production environments...
04/08/2026
Company will showcase updates across PRIME, Chyron LIVE, Paint, virtual production, newsroom workflows, and venue control...
04/08/2026
Exhibit will feature SilverBack, VersaBrix, VersaFrame, VF-9100, and FiberSaver-10G platforms...
04/08/2026
Platform updates target cloud-based production, high-density ST 2110 monitoring,...
04/08/2026
Live Media Group will take control of Gravity Media's U.S. mobile-unit fleet...
04/08/2026
Amp-modelling rigs get major upgrade
HeadRush have recently released a major firmware update that brings some powerful new features to their Prime, Core and...
04/08/2026
Generate a choir from a single voice track
Following years of development, Am I Audio have just officially launched their debut plug-in. Created in collabor...
04/08/2026
Official EMT-endorsed reverbs now available to all
Two more of Universal Audio's renowned UAD plug-ins have just gained new native versions that can be ...
04/08/2026
SBS helps Australians understand why every Census counts
Media releases
As Australia prepares for the 2026 Census on Tuesday 11 August, SBS is helping ensur...
04/08/2026
Canada invests in the future of HF communications with Rohde & Schwarz solutions After selection by the Canadian Department of National Defence (DND) as the p...
04/08/2026
Blendkit Expands Beyond Blender
Brie Clayton August 4, 2026
0 Comments
Popular 3D Asset Library Launches Alpha Plugins for Maya, Godot, and Rhino
Blend...
04/08/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...