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...
04/08/2026
Dalet, a leading technology and service provider for media-rich organizations, t...
15/07/2026
The 2026 FIFA World Cup is a big deal for DAZN, which has distribution rights in three markets: Italy, Spain, and Japan. Each requires a different editorial app...
15/07/2026
As sports-media organizations manage rapidly growing content libraries and incre...
15/07/2026
HBS director of digital Johannes Franken talks SVG Europe through the operation built to serve FIFA World Cup 2026 rights holders remotely from Stratford, east ...
15/07/2026
IBC has unveiled the conference program for IBC2026, with this year's event ...
15/07/2026
Software-defined wideband RF platform supports 96-kHz monitoring, streamlined fr...
15/07/2026
New AI-powered module enables broadcasters and content owners to extract dialogu...
15/07/2026
Weeklong gathering at EBU headquarters will test new IPMX capabilities and expand certified product ecosystem...
15/07/2026
Force Tech has announced the appointment of Bryan Adams, CTS, as Vice President of Integration. Adams will oversee AV and broadcast integration project delivery...
15/07/2026
USA Sports has announced an exclusive English-language agreement with the Bundes...
15/07/2026
ITV Sport's FIFA World Cup 2026 coverage has been produced from one of the m...
15/07/2026
Program will provide professional camera kits, training, and technical support to participating local newsrooms...
15/07/2026
San Francisco church becomes first U.S. deployment of the 2 M/E production switcher...
15/07/2026
Former ESPN and NFL Network executive to advise startup league on media, distribution, technology, and fan engagement...
15/07/2026
The Houston-based RSN is the live-game home of the Astros and Rockets...
15/07/2026
Launched alongside new FreqSoft installer
Designed to bridge the gap between hardware and software, Freqport's FreqInOut FO1 unit provides a simple way ...
15/07/2026
Save up to 55% on software until 23 July 2026
Expressive E have just launched their Summer Offer sale, with discounts of up to 55% now applied across their ...
15/07/2026
Real-time stem-separation tool updated
zplane have recently released a new and improved version of their real-time stem-separation plug-in which they say ma...
15/07/2026
A new benchmark in bass enhancement
Described as a new benchmark in bass enhancement , Baby Audio's latest release has been designed to augment the lo...
15/07/2026
DMC Production's AR 1 and AR 2 units reverse the conventional OB workflow, keeping all production-critical processing at the venue while operators work remo...
15/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
15/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
15/07/2026
GigaCloud, a leading Ukrainian cloud service provider, and Cubbit, the first geo-distributed cloud storage enabler, today announced a new Business Alliance Part...
15/07/2026
SMPTE, the global society for media professionals, technologists, and engineers, has announced the launch of the SMPTE 110th Anniversary Campaign, a fundraising...
15/07/2026
Studio Technologies, a leading manufacturer of high-quality audio, video, and fiber-optic solutions, unveils the new StudioComm Model 762 Stereo Level Control, ...
15/07/2026
COW Jobs: Producer, Virtual Events - Hybrid, Full Time, Salaried
Brie Clayton July 15, 2026
0 Comments
Producer, Virtual Events (Hybrid | Full-time | ...
15/07/2026
Brazilian media organization builds nationwide CDN with Broadpeak's high-performance software for low-latency streaming and new content distribution service...
15/07/2026
DHD announces a new utility designed to increase the efficiency of its audio mixing systems during initial installation and subsequent configuration.
Anyone w...
15/07/2026
BCNEXXT Brings Next-Generation Vipe to IBC2026, Breaking the 1:1 Channel-to-Server Ratio New innovations liberate broadcasters from legacy playout infrastructur...
15/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
15/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
15/07/2026
Wednesday 15 July 2026
Screeners and press pack now available: Gomorrah: The Or...
15/07/2026
Trinity's Treasures features Ruth Negga, Eleanor McEvoy, and David Norris, a...
15/07/2026
Join geographer Dr Susan Hegarty, Engineer Tim Joyce and Architect Orla Murphy a...
15/07/2026
Home to leading manufacturers, robotics pioneers, infrastructure builders and gaming, of course, Japan is one of the world's centers of AI - building across...
14/07/2026
South end zone videoboard, cloud-based control system will be ready for 2026 football season...
14/07/2026
Redesigned website, enhanced mobile app unify content, ticketing, and personalized fan engagement...
14/07/2026
Agreement spans sponsorship sales, digital monetization, radio production, and new practice facility naming rights...
14/07/2026
Independent league adds 14 regional sports network affiliates, growing distribut...
14/07/2026
New strategy aims to unify competitions, content, fan engagement, and commercial...
14/07/2026
ESPN and Disney+ become exclusive home of PFL events in key international MMA market...
14/07/2026
Gill will oversee engineering, technology, and sports operations across the company's 64 local television stations...
14/07/2026
Submitted by North American Broadcasters Association (NABA)
As broadcasters con...
14/07/2026
Modernized mobile unit combines HOME Apps, mc 56 console, and IP infrastructure ...
14/07/2026
Every match of the 32-team tournament will air across ION and Scripps Sports platforms in English and Spanish...
14/07/2026
FloHockey to stream every game of the annual international tournament under four...
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.
