100 models across 8 providers — real pricing, context windows, and what each is best for.
Anthropics fastest and most compact model. Ideal for near-instant responsiveness and high-volume tasks where cost matters.
Anthropics most powerful Claude 3 model. Excels at complex analysis, research, and nuanced understanding. Superseded by Claude 3.5 models.
Balanced performance and speed from the Claude 3 generation. Superseded by Claude 3.5 Sonnet but still available.
The fastest Claude 3.5 model. Matches Claude 3 Opus intelligence at a fraction of the cost, with near-instant response times.
Anthropics flagship model. Outperforms Claude 3 Opus at twice the speed and a fraction of the cost. Exceptional at coding, writing, and analysis.
Anthropics most intelligent model to date. Features extended thinking mode for complex multi-step reasoning and hard problems.
Fastest, most cost-efficient model
Fast, lightweight tasks and high-throughput applications
Most intelligent model for agents and coding
Most capable tasks — complex reasoning, research, and long-form writing
Legacy high-intelligence model
Most capable tasks — complex reasoning, research, and long-form writing
Anthropic's most powerful model, excelling at complex reasoning, nuanced analysis, and demanding tasks that require depth and precision.
Most capable tasks — complex reasoning, research, and long-form writing
Optimal balance of intelligence, cost, and speed
Legacy model with balanced performance
Balanced performance and speed for production applications
DeepSeeks fast, efficient, and economical model. 284B total / 13B active params with 1M context. Supports both non-thinking and thinking modes.
DeepSeeks frontier open-source model. 1.6T total / 49B active params with 1M context. Rivals top closed-source models in coding, math, and STEM reasoning.
Googles fast, efficient multimodal model with a massive 1M token context window. Excellent price-to-performance for most tasks.
Googles most capable Gemini 1.5 model with a 2M token context window — the largest of any commercially available model. Excellent for long documents and video.
Next-generation fast model from Google. Improved performance over 1.5 Flash with added agentic capabilities and real-time information access.
Gemini 2.0 Flash-Lite
Googles most capable Gemini 2.0 model. Strongest coding and world knowledge of the Gemini family. Designed for complex agentic tasks.
Low-latency, high-volume tasks that require reasoning with best price-performance.
The fastest and most budget-friendly multimodal model in the 2.5 family.
Complex tasks featuring deep reasoning and coding capabilities.
Efficient processing of multimodal content with fast inference.
Frontier-class performance rivaling larger models at a fraction of the cost.
Complex multimodal tasks and advanced reasoning
Fast multimodal processing with image generation capabilities
Advanced intelligence, complex problem-solving, and powerful agentic and vibe coding capabilities.
Most intelligent model for sustained frontier performance on agentic and coding tasks.
Autonomous multi-step research across hundreds of sources to produce cited, interactive reports.
Maximum comprehensiveness for automated context gathering and synthesis across hundreds of sources.
Advanced embodied reasoning model for robotic agents with spatial and physical reasoning.
High-quality image generation and editing
High-quality music generation from text prompts
Advanced video generation from text and image prompts
Text summarization, multilingual agents, and coding use cases with tool use capabilities.
Metas largest and most capable open-source model. Competes directly with GPT-4o and Claude 3.5 Sonnet on many benchmarks — and its free.
Text summarization, multilingual agents, and coding use cases with maximum capability.
Text summarization, multilingual agents, and coding use cases.
Metas mid-size open-source model. Competitive with many commercial models at zero licensing cost. Requires significant hardware to self-host.
Metas smallest open-source Llama 3.1 model. Free to download and run locally. Surprisingly capable for its size, especially for on-device use.
Text summarization, multilingual agents, and coding use cases.
Cost-effective edge deployment with multimodal capabilities for image reasoning.
Flexible multimodal reasoning on high-resolution images with text output.
Metas first open-source multimodal Llama model. Understands images and text together. Ideal for vision tasks without cloud dependency.
Lightweight, cost-efficient edge deployment for text-based applications.
Lightweight, cost-efficient edge deployment for text-based applications.
Flexible multimodal reasoning on high-resolution images with text output.
Multilingual text-based use cases such as synthetic data generation.
Text-based use cases such as synthetic data generation with 405B-level performance.
Metas latest 70B model with improved instruction following and reasoning. Delivers 405B-level performance in a more efficient package.
Long-form multimodal work with native image and text understanding.
Long document analysis with efficient single GPU deployment.
Preview reasoning-enabled chat model with built-in problem-solving for conversational interfaces.
Maximum reasoning effort code model with extended capabilities.
Lightweight code-focused model for developer tooling.
Conversational AI with structured outputs and tool calling support.
Latest preview chat model with reasoning capabilities and structured outputs for general conversational AI tasks.
Microsofts consumer and enterprise AI assistant powered by GPT-4o. Deeply integrated into Windows, Microsoft 365, Edge, and Bing.
Microsofts small but mighty language model. Designed for on-device and edge deployment. Punches well above its weight class on reasoning tasks.
Video generation from text instructions.
General-purpose language tasks
General-purpose language tasks
Previous generation flagship from OpenAI. Largely superseded by GPT-4o but still available. Solid performance on complex reasoning tasks.
Instruction-following and agentic coding workflows
Instruction-following and agentic coding workflows
Instruction-following and agentic coding workflows
OpenAIs flagship multimodal model. Fast, highly capable across text, vision, and audio. The most widely used frontier model.
A smaller, cheaper version of GPT-4o. Surprisingly capable for its price point. Ideal for high-volume, cost-sensitive applications.
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
General-purpose language tasks
OpenAIs first reasoning model. Thinks before answering using chain-of-thought internally. Excels at math, science, and complex coding problems.
Deep reasoning tasks requiring careful step-by-step thinking
Advanced reasoning and complex problem-solving
Advanced reasoning and complex problem-solving
Advanced reasoning and complex problem-solving
Perplexitys lightweight search-augmented model. Every response is grounded in live web search results with citations. Fast and cost-effective.
Perplexitys most powerful search model. Deeper research capability with more search queries per response and stronger reasoning over retrieved information.
xAIs capable conversational model with real-time internet access baked in by default. Deep integration with X (Twitter) data and current events.
xAIs most advanced model. Significant leap in reasoning and intelligence. Trained on a massive compute cluster and competitive with frontier models.
Agentic tool calling tasks requiring low hallucination rates, strict prompt adherence, and fast response times.
Recommended for non-reasoning workloads as a replacement for deprecated models.
Delivering truthful and insightful answers with a focus on truth-seeking capabilities.
Fast coding model trained specifically for agentic coding workflows.