
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 ...
09/10/2026
September 10 2026, 06:00 (PDT) Dolby Expands Dolby OptiView Platform with New Capabilities at IBC 2026
New Sports Intelligence helps providers better unders...
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...
10/09/2026
France launch follows UK and Germany as Nielsen expands product across EMEA
Paris, France, September 10, 2026 - Nielsen, a global leader in audience measuremen...
10/09/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
10/09/2026
Leading professional audio manufacturers NUGEN Audio, RTW, Steinberg and Telos Alliance have joined forces with Fraunhofer IDMT to establish a common approach t...
10/09/2026
Built Into Setplex's Zapflex Platform, AI Insider Turns Video Service Data Into Actionable Insights
Setplex today introduced AI Insider, a breakthrough new...
10/09/2026
Managed service changes dialogue without actor-specific training or facial rigging while preserving the original performance and fitting established post-produc...
10/09/2026
The ACT Entertainment tvONE and Green Hippo Pixel Academy has released its autumn course schedule, with hands-on training dates in September, October and Novemb...
10/09/2026
Orange Polska continues to Rely on Viaccess-Orca for Secure TV Services and the Next Phase of its Technology Evolution
Viaccess-Orca (VO), a global leader in T...
10/09/2026
Generate and create directly in your timeline with new AI-powered innovations in...
10/09/2026
Tedial is introducing Media Business Hub (MBH) at IBC 2026, a new business enablement platform designed to help media organizations unlock the full value of the...
10/09/2026
Globecast, a leading provider of broadcast, media and entertainment managed services, will unveil the next phase of its transformation at IBC2026 with a preview...
10/09/2026
Fabric Data and TMT Insights today announced that the two companies are combining to create Fabric Insights, uniting authoritative data, purpose-built software,...
10/09/2026
Operative, the preferred advertising and broadcast management provider for the world's leading media brands, today announced a comprehensive slate of new pr...
10/09/2026
Test & measurement innovator, Leader Electronics of Europe, has announced the launch of its latest range of handheld signal generation, analysis and monitoring ...
10/09/2026
The partnership begins with native Zixi protocol support on the Kiloview D350 and will extend to deeper product integration and joint go-to-market initiatives
...
10/09/2026
Avid and FilmLight announce strategic partnership to reimagine the color pipelin...
10/09/2026
Maxon Brings Exclusive Preview of Fall Innovations to IBC2026
Brie Clayton September 10, 2026
0 Comments
IBC attendees will get the first look at new ...
10/09/2026
Puget Systems Launches Built Different by Puget Systems, A New Podcast That Go...
10/09/2026
Three Boston Conservatory Dance Alums Honored for Their Early-Career Promise India Hobbs and Demetrius Lee received Princess Grace Awards, while Lilly David t...
10/09/2026
MonoNeon Returns to Berklee to Kick Off Signature Series The Grammy-winning bassist and alumnus performs with students and faculty from the Planet MicroJam In...
10/09/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
10/09/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
10/09/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
10/09/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
10/09/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
10/09/2026
Harmonic Leads the New Era of Broadband Connectivity at SCTE TechExpo 2026 Showcase Spotlights Include:
New cOS SensAI intelligence fabric unleashing data-pow...
10/09/2026
Agility Beyond Space and Levira partner to expand video distribution across Eura...
10/09/2026
RT to Broadcast Amgen Irish Open free-to-air
New multi-year agreement will see RT continue to bring the Amgen Irish Open to Irish audiences
Ryder Cup hig...
10/09/2026
More than 30 women report allegations of sexual harassment, misogyny and bullying within National Ambulance Service
Tonight on RT Prime Time, 9:35pm, RT One ...
10/09/2026
Series 3 of Nero's Class begins Friday 11 September on RT Listen
The third...
10/09/2026
Gear up: The latest PC games and major updates are ready to play on GeForce NOW ...
10/09/2026
AI inference chipmaker d-Matrix today announced it will use NVIDIA NVLink Fusion to connect its next-generation Raptor XPUs to NVIDIA's AI infrastructure pl...
09/09/2026
Sony Electronics has announced the SEL814G, the company's first fisheye zoom G Lens for full-frame Alpha E-mount cameras. The lens will be available in Octo...
09/09/2026
Telestream has announced an expansion of its Practical AI capabilities across th...
09/09/2026
Haivision and Chyron have announced a technology partnership integrating Haivisi...
09/09/2026
Omaha Productions has released ManningCast: Heist, a short film featuring Peyton and Eli Manning attempting to break into NFL Commissioner Roger Goodell's o...
09/09/2026
Imagine Communications will demonstrate the next evolution of XVR at IBC2026 (Stand 1.B73, RAI Amsterdam, September 11-14).
XVR is Imagine's channel-in-a-b...
09/09/2026
swXtch.io has announced abbe, an AI-based orchestration layer for live media workflows, making its world premiere at IBC2026 (Meeting Room 5.MR13, Hall 5, Septe...
09/09/2026
NFL Films has announced the second season of Inside the NFL on X as an X Original, premiering Wednesday, September 9. The series marks Inside the NFL's 50th...
09/09/2026
LiveU has announced an expanded partnership with The Associated Press, deploying the largest fleet of LU900Q intelligent production units in the industry as par...
09/09/2026
Diversified has announced plans to consolidate its four existing New Jersey locations into a new facility in Newark, expected to be fully operational by the end...
09/09/2026
Wisycom will exhibit new RF products at IBC2026 (Stand 8.D60, September 11-14), including the MATF Wideband Antenna Matrix, PFL Portable RF over Fiber Box, and ...
09/09/2026
Guntermann & Drunck (G&D) and VuWall, both Panoptec companies, will introduce vsNEO at IBC2026 (Stand 8.B63, September 11, RAI Amsterdam). Pre-orders open in ea...
09/09/2026
Mo-Sys has completed a business restructuring and relocated to a new facility in Canada Water, London. As part of the reorganization, Selin Kemal has been promo...
09/09/2026
Aicox has integrated TAG Video Systems' Core monitoring platform into Telef nica's TSAmediaHUB audiovisual contribution and distribution infrastructure....
09/09/2026
Blackmagic Design has released DaVinci Resolve 21.1, a free update available now through the Blackmagic Design website. The release adds new AI tools, workflow ...
09/09/2026
ESPN and Apple Music have announced a collaboration on music curation for Monday...
09/09/2026
Sacramento Production Services has completed a wireless audio overhaul at Sutter Health Park, home of the Athletics during their three-year residency in Sacrame...