Top news
Investing in the public benefit OfficialAnthropic NewsOct 18Popular Musk Pledges SpaceX Shares to ‘Every Kid in America’ WebBloomberg TechnologyOct 19Popular Websites are closing their doors to AI agents WebThe Verge AIOct 17 This is the new ChatGPT Voice, powered by GPT Live OfficialOpenAI on YouTubeOct 18 How AI giants will reshape global philanthropy WebAxios AIOct 19 Gemini 4 Argon Really Surprised Me WebTwo Minute PapersOct 19 Google Gemini 4 agents just beat Minecraft with no human intervention WebEngadget AIOct 19 Who could it be? 🧐 OfficialGoogle on YouTubeOct 18 Anthropic gives outside AI watchdog “employee-like” access WebArs Technica AIOct 18 What are self-replicating ‘AI worms’, and why is Anthropic worried about them? WebThe Guardian AIOct 18
GlossaryModelsIntermediate

Parameters

Also written: Model scale, Model size, Parameter Golf, Trillion Parameter Model, Trillion-parameter, Parameter signatures, Function signature, Method signature

Parameters are the learned internal variables of an AI model that determine its behavior and capabilities.

In plain terms

Parameters are the core numbers inside an AI model that it learns during training. Think of them like the knobs and dials on a complex machine that are adjusted to get the desired output.

During training, the AI looks at vast amounts of data and adjusts these parameters to get better at its task, whether that's understanding text, generating images, or making predictions. The more parameters a model has, the more complex patterns it can potentially learn, but it also requires more data and computational power to train.

The expert version

In the context of neural networks, parameters are the weights and biases that the model learns from training data. These values are adjusted through optimization algorithms (like gradient descent) to minimize a loss function, thereby enabling the model to perform its intended task.

The number of parameters is often used as a proxy for model size and potential capability. For instance, models like GPT-3 have billions of parameters, while newer models can have trillions. However, a higher parameter count does not automatically guarantee better performance; factors like training data quality, architecture, and training techniques are also crucial. Tradeoffs involve computational cost for training and inference, memory requirements, and the risk of overfitting.

Why it matters now

Recent news highlights the ongoing race to develop larger and more capable AI models, with mentions of "trillion parameter" models suggesting a new scale of complexity. Companies are also refining how developers interact with these models, as seen in updates to API cookbooks that fix "SDK parameter signatures." This indicates a focus on both the internal scale of models and the external interfaces for using them.

Where it shows up 18 items, newest first
How each company uses it
Anthropic Claude
Anthropic's research on model persuasiveness implies that larger models, often indicated by more parameters, lead to improved performance in areas like Claude's generation capabilities.
Google Gemini
Google Gemini's API Cookbook shows a focus on managing and updating "SDK parameter signatures," which are the specific inputs and outputs developers use when interacting with their models.
Meta Meta AI
Meta AI's open-weight coding model suggests that the scale, or number of parameters, is a key factor in its ability to be customized for AI agent development.
OpenAI ChatGPT
OpenAI is mentioned in the context of "trillion parameter" models, indicating their involvement in developing or discussing models at this advanced scale.
Related terms

Neural NetworksWeightsBiasesTraining DataModel SizeGradient Descent

Often confused with
Hyperparameters
These are settings that are not learned from data but are set b
Features
These are the input variables or characteristics of the data that the
Neurons
These are the individual computational units within a neural network; p
Origin

The concept of parameters originates from the field of statistics and machine learning, referring to the coefficients or values that define a statistical model. In neural networks, parameters specifically refer to the weights and biases learned during training. Source ↗

Try this about 5 minutes

You can observe the effect of parameters by comparing different model sizes if available. For example, try asking a simpler chatbot model (likely with fewer parameters) and a more advanced one (with more parameters) the same complex question. Notice how the more advanced model might provide a more nuanced, detailed, or accurate answer, reflecting its greater capacity learned through its parameters.

Try it: Playbook entries that use thisPlaybook →

The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.

Questions people ask

How many parameters does a typical AI model have?
The number varies greatly, from thousands in simple models to billions or even trillions in large frontier models like GPT-4.
Do more parameters always mean a better AI?
Not necessarily. While more parameters often lead to greater capability, they also require more data, computational power, and can be harder to train effectively without overfitting.

First seen on Subvolts Sep 22, 2020. Updated Oct 17, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.