Frontier intelligence, unified.

Anthropic's premier model for nuanced reasoning, software engineering, and high-precision document synthesis.

Anthropic's pinnacle intelligence tier leading global benchmarks in autonomous software engineering and formal verification.

Enterprise workhorse balancing Opus-class reasoning with fast token generation and hybrid configurable thinking budgets.

Open-weights reasoning model utilizing large-scale reinforcement learning for complex step-by-step proofs and code.

671B parameter Mixture-of-Experts (MoE) activating 37B parameters per token using Multi-Head Latent Attention (MLA).

Google frontier model featuring industry-leading 2M token context window for massive repos, long docs, and media.

Google's high-capacity reasoning model featuring 2M tokens context for large-scale code repositories and video archives.

Google DeepMind's workhorse multimodal foundation model with 1M context, configurable thinking budgets, and native video.

Real-time conversational speech-to-speech engine with live extended thinking, acoustic interruption, and sub-200ms latency.

OpenAI flagship omni model with high-speed intelligence across text, vision, and structured schema generation.

OpenAI's frontier agentic model engineered for autonomous computer use, desktop GUI interaction, and repository refactoring.

OpenAI's cost-efficient enterprise foundation model delivering frontier intelligence at 70% reduced inference expenditure.

Real-time bidirectional speech-to-speech audio API streaming continuous sub-300ms natural conversational audio.

xAI's frontier reasoning engine trained on the Colossus supercluster with live real-time web grounding and coding prowess.

High-performance Ling 3.0 Flash VL for coding and reasoning.

Alibaba's open-weights code generation state-of-the-art model trained on 5.5T tokens across 92 programming languages.

Anthropic's specialized creative intelligence model fine-tuned for expressive narrative prose, storytelling, and stylistic voice.

Ultra-fast sub-150ms inference model generating 140+ tokens per second for high-concurrency enterprise automation.

Mistral AI's state-of-the-art coding and FIM (Fill-in-the-Middle) model optimized for low-latency developer copilot tools.

Enterprise RAG and multi-step tool use model specialized in citation grounding and verifiable multilingual retrieval.

High-throughput, cost-effective multimodal model with a 1M token context window built for rapid response times.

Google's hybrid reasoning engine with dynamic token thinking allocation and 1M context multimodal understanding.

Sub-cent high-throughput multimodal model engineered for high-concurrency ingestion and rapid classification.

Native omnimodal model processing interleaved image, audio waveforms, video frames, and text synchronously.

Affordable and exceptionally fast lightweight model for everyday tasks, triage, and high-frequency queries.

Intermediate enterprise frontier model offering dependable tool calling, code generation, and low hallucination rates.

Massive scale knowledge and factual retrieval model designed for analytical research and structured data transformation.

Ultra-fast, high-throughput micro-task model with sub-180ms TTFT for semantic routing, classification, and real-time chat.

xAI's high-speed intelligence engine with built-in real-time search synthesis and vision comprehension.

Meta's flagship open-weights foundation model serving as the global standard for synthetic distillation and complex reasoning.

Meta's most capable open-weights instruction model, rivaling closed models on code generation, reasoning, and math.

Microsoft's 14B state-of-the-art small language model rivaling 70B models in mathematical reasoning and synthetic data quality.

Flagship European model specialized in advanced multilingual reasoning, multi-turn dialogue, and function calling.

High-density lightweight foundation model designed for edge devices and fast serverless microservice invocation.

NVIDIA's customized Llama model optimized with reward model training to achieve leading Arena Elo scores and helpfulness.