The Archive

OpenAI API Structured Outputs enforce developer-supplied JSON schema compliance in model responses.

Anthropic makes Claude accessible in Brazil, expanding geographic availability for regional users.

OpenAI primer on EU AI Act compliance: deadlines, prohibited/high-risk use cases for AI providers and deployers.

SearchGPT prototype: new search features delivering fast, sourced answers integrated with AI capabilities.

Mistral Large 2 flagship release with improved code, math, reasoning, multilingual support, and function calling.

Rule-Based Rewards method aligns language models to safe behavior without requiring extensive human feedback data.

GPT-4o mini: cost-efficient model advancing accessible AI deployment.

Mistral AI releases Mistral NeMo, a 12B open-weights model with 128k context, built with NVIDIA under Apache 2.0.

ChatGPT Enterprise: Compliance API, SCIM, and GPT controls for enterprise data security and access governance.

Anthropic partners with Menlo Ventures to launch Anthology Fund for AI startup investing and ecosystem development.

Prover-verifier games improve legibility and verifiability of language model outputs for human and machine trust.

Mistral AI releases Codestral Mamba, a Mamba2-based code generation model under Apache 2.0 license.

Mistral AI releases Mathstral, a 7B open-weights model specializing in mathematical reasoning and scientific tasks.

Anthropic enables fine-tuning of Claude 3 Haiku via Amazon Bedrock for custom enterprise deployments.

OpenAI and Los Alamos National Laboratory develop safety evaluations for biological capabilities and risks of frontier models.