Latest LLM

Agentic Coding, Large-scale Refactoring, Cybersecurity Defense, Codex CLI/IDE integration, Long-horizon tasks.

Unified Reasoning, 400k Context, Agentic Workflows, Native Multimodality, High-Reliability.

A state-of-the-art, open-weight, general-purpose model.

A powerful model offering best-in-class text generation capabilities.

A powerful and efficient model offering best-in-class performance for its size.

A tiny and efficient model offering best-in-class performance on edge devices.

Anthropic's most powerful model (4.5 series), excelling in complex tasks, reasoning, and creativity.

New generation of Google Gemini, natively multimodal with increased reasoning capabilities.

Iterative improvement of GPT-5, offering better reliability and more nuanced responses.

Advanced Moonshot AI model with reinforced 'Thinking' capabilities.

Performance/cost balance of the 4.5 series, ideal for enterprise and scaling.

Our frontier-class multimodal reasoning model.

Fast, efficient Claude 4.5 series model optimized for low latency and cost.

Our frontier-class multimodal model released for general availability.

OpenAI's next major leap, promising general intelligence closer to human level.

Intermediate update to the Claude 4 Opus series.

Our cutting-edge language model for coding tasks.

A mini version of our first audio input model.

Our first model with audio input capabilities for general use.

Performant mid-sized open-source model (20B), optimized for self-hosting.

Large open-source model (120B) rivaling proprietary models in reasoning.

An enterprise grade text model, that excels at development tasks.

An update to our open source model that specializes in development.

Fourth iteration of xAI, deeply integrated with real-time data.

An efficient audio input model, fine-tuned for transcription tasks.

An update to our previous small model, optimized for efficiency.

The 'workhorse' of series 4, efficient and versatile.

Our state-of-the-art semantic model for extracting embeddings from code.

Flagship model of generation 4, pushing boundaries of context and understanding.

Ultra-fast and economical model from Google, ideal for high-frequency applications.

Flagship, Native Multimodal, MoE (400B), Advanced Reasoning, SOTA, High-Performance.

Our frontier-class multimodal model released for general availability.

Our OCR service powering our Document AI capabilities.

Small, performant reasoning model, economical successor.

Very compact model for embedded or mobile applications.

Advanced generalist model, text and code, successor to GPT-4/Turbo.

Efficiency, Speed, Optimized MoE, Low-Latency, High-Reasoning/Low-Cost, Agile.

Miniaturized and optimized version of the GPT-4.1 branch.

Second generation xAI Grok model, oriented towards conversation, code, and real-time web access via X (Twitter).

Our moderation service that enables our safety capabilities.

Fast and light version of Claude 3.5, optimized for latency and cost.

Development assistant based on OpenAI models (including GPT-4o), optimized by GitHub for code, reviews, and context generation.

Reasoning model optimized for complex tasks, logic, and problem solving (advanced reasoning).

Large Perplexity model based on Llama 3.1, specialized in online search and cited answers.

Lighter and faster version of Perplexity sonar-online, adapted for real-time usage.

Intermediate Claude 3.5 model, great speed/quality balance, strong in reasoning and code.

Real-time multimodal model (text, image, audio) optimized for interactivity and cost.