Model Marketplace

Most capable multimodal model — vision, audio, tool use, and streaming.

Fast, affordable GPT-4 class model — great for high-volume tasks.

Most capable OpenAI embedding model for semantic search and classification.

Anthropic's most intelligent model — coding, analysis, and long-context reasoning.

Fastest Anthropic model — near-instant responses for lightweight tasks.

Google's next-gen fast multimodal model — 1M context, vision, and tool use.

OpenAI open-source 120B running on Groq LPU — 500 t/s, 131K context.

OpenAI open-source 20B on Groq — blazing 1000 t/s output speed.

Safety-focused 20B model for content moderation and prompt injection filtering.

Alibaba's Qwen 3.6 27B on Groq — strong at coding and multilingual tasks.

Alibaba's Qwen 3.8 27B on Groq — enhanced reasoning and instruction following.

Agentic AI system with built-in web search and code execution at 450 t/s.

Lightweight Compound — fast agentic responses with integrated web search.

OpenAI Whisper Large V3 on Groq — speech-to-text at 164× real-time speed.

Faster Whisper variant on Groq — audio transcription at 400K audio seconds/hour.

Tiny 22M parameter model for prompt injection and jailbreak detection.

86M parameter prompt guard — more accurate injection and jailbreak detection.

State-of-the-art reasoning model with transparent chain-of-thought.

Mistral's flagship model — complex reasoning, coding, and long-context tasks.

NVIDIA's 340B parameter model optimised for enterprise AI workloads.

Moonshot AI's next-gen reasoning model hosted on NVIDIA NIM with TensorRT-LLM acceleration.

DeepSeek's flagship reasoning & coding powerhouse on NVIDIA NIM with TensorRT-LLM.

Ultra-fast, low-latency DeepSeek model hosted on NVIDIA NIM for high-throughput applications.

NVIDIA's 1B parameter embedding model for high-accuracy semantic search and enterprise RAG.

Moonshot AI's long-context model — 200K token window with vision and reasoning.

Cohere's most powerful model for complex RAG workflows and tool use.