Leading AI models,
one production API.

Frontier model for complex reasoning, coding and agent execution.

Balanced frontier intelligence with faster production throughput.

Efficient reasoning for high-volume agentic applications.

High-accuracy general intelligence for demanding production work.

Fast multimodal reasoning with a lower cost profile.

Compact model for classification, extraction and routing.

Reliable everyday intelligence at high throughput.

The lowest-cost OpenAI route for focused tasks.

State-of-the-art image generation and editing through the Images API.

Realtime speech model with reasoning and native tool use.

Lower-cost realtime voice route with reasoning and tool support.

Production speech-to-speech model for responsive voice agents.

Streaming speech-to-speech translation for multilingual experiences.

Streaming speech recognition for live transcription workflows.

High-accuracy speech-to-text model for recorded audio.

Fast, economical transcription for high-volume audio.

High-quality vector representations for retrieval, clustering and RAG.

Cost-efficient embeddings for semantic search and classification.

Advanced reasoning and coding with strong agent reliability.

Anthropic's highest-capability route for complex knowledge work.

Deep reasoning for long-running engineering and research tasks.

Reliable high-depth analysis and agentic coding.

High-capability Claude route for long-running coding and enterprise tasks.

Deep reasoning model for complex engineering and research workflows.

Balanced intelligence for coding and business workflows.

Reliable coding and analysis model with broad cloud availability.

Fast Claude model for responsive user-facing experiences.

Native multimodal reasoning with a one-million-token context.

High-depth preview model for complex multimodal work.

Fast preview route for high-volume multimodal applications.

Cost-efficient Gemini route for extraction and classification.

Stable price-performance model for low-latency multimodal reasoning.

Fastest budget Gemini route for high-throughput multimodal work.

High-efficiency native image generation and editing.

Ultra-low-latency image generation for interactive applications.

High-context native image generation and editing for premium creative work.

Low-latency Live API model for voice-first multimodal applications.

Unified embeddings for text, image, video, audio and document retrieval.

Large multimodal open-weight model for general workloads.

Efficient multimodal Llama route with broad context.

Production-tuned open model for chat and agent workflows.

Fine-tunable compact Llama model for specialized private workloads.

Large open multimodal model for image and document understanding.

Efficient open vision model for private multimodal applications.

Large open-weight foundation model for demanding inference workloads.

Ultra-fast open route for simple high-volume tasks.

Mistral's production generalist for multilingual work.

Deep reasoning with efficient European deployment options.

Compact production model for agents and structured output.

Open text-and-vision model for capable private deployments.

Efficient open model for edge and private deployments.

Small open-weight route for focused automations.

Frontier code-agent model for repository-scale software engineering.

Code completion model optimized for low-latency developer tooling.

Multilingual text-to-speech with zero-shot voice cloning.

Open realtime transcription model for streaming audio.

Efficient transcription model for recorded speech.

Document parsing with structural blocks and paragraph-level coordinates.

Safety classification with jailbreak and policy-risk detection.

Semantic vectors for multilingual retrieval and RAG.

Qwen's flagship reasoning and coding model.

Long-context multilingual reasoning at efficient cost.

Fast multimodal Qwen route for interactive products.

Open mixture-of-experts model for reasoning and code.

Large open mixture-of-experts model for math, code and agents.

Efficient open MoE route for local reasoning and tool use.

Open vision-language model for documents, images and visual reasoning.

Multimodal Moonshot model for reasoning, coding and agent workflows.

Moonshot's flagship model for long-horizon agent work.

Specialized coding model for repository-scale tasks.

Balanced agentic reasoning with reliable tool use.

Long-horizon reasoning and software engineering model.

Low-latency GLM route for interactive agents.

Fast general reasoning and multilingual tool use.

Agentic engineering model for sustained autonomous software work.

Open-weight foundation model for long-horizon engineering agents.

Multimodal coding model specialized for visual programming.

Stable multi-step reasoning and agentic coding model.

Lightweight high-speed GLM route for interactive agent applications.

Vision-language model with native function calling and thinking control.

Open image generation model for complex prompts and text rendering.

Document parsing and structured information extraction model.

xAI's flagship reasoning model for agents, code and multimodal work.

Coding-focused Grok route for repository work and software agents.

High-throughput general reasoning with a one-million-token context.

Native multi-agent orchestration for parallel research and execution.

High-quality image generation and editing up to 2K resolution.

Fast image generation and editing for interactive products.

Image-to-video generation with 480p, 720p and 1080p outputs.

Text, image and video conditioned generation and video extension.

Realtime voice agent model with fast reasoning and tool use.

Natural speech synthesis through xAI's Text to Speech API.

Batch and streaming speech recognition across 25 languages.

Frontier-scale open reasoning model for complex agents and long-context code.

Efficient open MoE for reasoning, planning, coding and tool calling.

Omnimodal model that understands video, speech, images and text.

Full-duplex speech-to-speech model that listens and speaks simultaneously.

Low-latency streaming speech recognition for enterprise voice systems.

Expressive zero-shot text-to-speech from a short voice sample.

Physics-aware video world generation from text or image prompts.

Structured physical-world reasoning over video and image inputs.

Fast OCR with layout, table and document structure extraction.

Low-latency, steerable speech generation with expressive audio tags.

Realtime speech-to-speech translation across more than 70 languages.

Conversational video generation and multi-turn editing from text and images.

Cinematic video generation with synchronized audio and reference controls.

Lower-cost Veo route for production video generation and editing.