Terms

Model Distillation

Follow the latest coverage, related explainers and connected technology stories.

Latest coverage

CN
Mistral Unveils Le Chonk Model with Open Weights and One Trillion Parameters

Mistral Unveils Le Chonk Model with Open Weights and One Trillion Parameters

Mistral announced Mistral Large 4, known as Le Chonk, a one-trillion-parameter model available with open weights and customizable, with a particular focus on programming, cyber defense, and specialized sectors. The preview release comes amid growing competition over access to and control of advanced artificial intelligence models.

CN
Anthropic CEO Proposes Plan to Slow the AI Race

Anthropic CEO Proposes Plan to Slow the AI Race

Dario Amodei proposed three mechanisms to “slow the pace of progress” in artificial intelligence models, including independent monitors within companies, shared safety standards, and international coordination. Anthropic, for its part, is committing to providing expanded access to external evaluators, while the feasibility of the proposals and their potential conflict with competition laws remain open questions.

CN
Garry Tan Calls on U.S. AI Labs to Distill Frontier Models

Garry Tan Calls on U.S. AI Labs to Distill Frontier Models

Garry Tan, CEO of Y Combinator, believes that smaller U.S. labs developing open-weight models should be able to use model-distillation techniques on advanced American models, rather than leaving this path exclusively to Chinese labs. He frames this as a way to maintain a balance between closed commercial models and open models, while rejecting the use of stolen credentials or fraudulent methods.

CN
Anthropic Detects Nearly 200 Million Interactions in Campaigns to Extract AI Model Capabilities

Anthropic Detects Nearly 200 Million Interactions in Campaigns to Extract AI Model Capabilities

Anthropic said that five campaigns linked to Chinese artificial intelligence companies attempted to extract advanced capabilities from the Claude model through nearly 200 million interactions, including one campaign attributed to Alibaba and another to Moonshot AI. The company believes these methods could be used to produce training data for competing models, although the report presents the company’s allegations and the material provides no independent evidence supporting them.

CN
U.S. Agencies Accuse Six Chinese Companies of Extracting Billions of Tokens from Advanced AI Models

U.S. Agencies Accuse Six Chinese Companies of Extracting Billions of Tokens from Advanced AI Models

CISA, NSA, and the FBI said that six Chinese artificial intelligence companies carried out extensive operations to extract outputs from advanced U.S. models through billions of tokens and millions of requests since late 2024. The agencies recommend improving monitoring of API abuse and sharing indicators of these campaigns among companies.

CN
Anthropic Launches Claude Fable 5.1 and Mythos 5.1 with Lower Agent Costs and Stricter Model Protection

Anthropic Launches Claude Fable 5.1 and Mythos 5.1 with Lower Agent Costs and Stricter Model Protection

Anthropic has introduced the Claude Fable 5.1 and Claude Mythos 5.1 models for programming, scientific research, and long-running tasks, cutting the cost of reading cached storage by 75% and making Mythos available to users who pass specialized verification programs. The company has also imposed new restrictions on API accounts to counter model-distillation attacks.