Text & Chat Models

Anthropic's most powerful model. Exceptional at complex analysis, nuanced writing, math, and coding. Sets new benchmarks across evaluation suites.

Anthropic's balanced model offering excellent performance at a lower cost than Opus. Great all-rounder for production workloads.

DeepSeek's reasoning model trained with reinforcement learning. Performs chain-of-thought reasoning, rivaling OpenAI's o1 on math and science benchmarks.

DeepSeek's flagship 671B MoE model. Competitive with GPT-4o on many benchmarks. Exceptional at coding, math, and Chinese language tasks.

Google's most advanced thinking model with built-in reasoning capabilities. Excels at complex tasks requiring multi-step reasoning.

OpenAI's latest frontier model with improved reasoning, creativity, and instruction following. Significant improvements over GPT-4o.

OpenAI's most capable multimodal model. Accepts text and image inputs, produces text outputs. Excellent for complex reasoning, creative writing, and analysis.

xAI's most powerful model. Trained on massive compute with strong reasoning, humor, and real-time knowledge from X (Twitter).

Meta's latest 70B model delivering performance comparable to Llama 3.1 405B at a fraction of the cost. Excellent open-source option.

Meta's powerful Llama 4 Maverick model. A larger, more capable variant with strong reasoning, creative writing, and multilingual abilities.

OpenAI's reasoning model that thinks before answering. Uses chain-of-thought to solve complex math, science, and coding problems.

Anthropic's fastest and most affordable model. Ideal for high-volume tasks, customer support, and quick responses.

Previous generation balanced model from Anthropic. Still excellent for many tasks including coding, analysis, and creative writing.

Mistral's dedicated code model. Trained specifically for code generation, completion, and understanding across 80+ programming languages.

Cohere's efficient model optimized for RAG and tool use. Great balance of quality and cost for production deployments.

Cohere's flagship model for enterprise RAG applications. Excellent at retrieval-augmented generation, summarization, and multi-step tasks.

Databricks' open-source MoE model with 132B total parameters. Strong at enterprise tasks, SQL, and data-related queries.

Uncensored Mixtral fine-tune. Open-ended assistant without content restrictions for research purposes.

Google's fast, versatile multimodal model. Supports text, images, audio, and video inputs. Great balance of speed and capability.

Google's most cost-efficient model. Optimized for high-volume, lower-complexity tasks with excellent throughput.

Google's open-source 27B model. Strong performance in reasoning and text generation, built with Google's research expertise.

Compact open-source model from Google. Excellent for on-device deployment and resource-constrained environments.

Google's compact open model on Replicate. Efficient 9B model with strong general capabilities.

Small, fast, and affordable model from OpenAI. Great for lightweight tasks like classification, summarization, and simple Q&A.

xAI's previous flagship model. Known for its witty personality, strong reasoning, and ability to handle nuanced questions.

Smaller, faster version of Grok 3. Excellent for quick responses and lower-cost applications while maintaining strong capabilities.

Meta's largest open-source model. 405 billion parameters delivering frontier-class performance on reasoning, coding, and multilingual tasks.

Meta's highly capable 70B open-source model. Great balance of performance and efficiency for a wide range of tasks.

Meta's popular 70B model on Replicate. Strong all-around performance for chat, coding, and reasoning.

Meta's compact 8B model. Surprisingly capable for its size, perfect for fast inference, edge deployment, and cost-sensitive applications.

Efficient 8B Llama model on Replicate. Fast and affordable for straightforward tasks.

Compact multimodal Llama 3.2. Vision-language model for efficient image understanding and text generation.

Smallest Llama model for on-device inference. 1B parameters, ideal for mobile and IoT applications.

Ultra-compact Llama model for edge deployment. 3B parameters with surprising capability for its size.

Meta's multimodal Llama 3.2. 90B parameter model with native image understanding and text generation.

Meta's next-generation Llama 4 model optimized for efficiency. Built on a new architecture with improved reasoning and instruction following.

Mistral's most capable model. 123B parameters with strong reasoning, multilingual support, and function calling. Great for complex enterprise tasks.

Mid-range model from Mistral AI offering a good balance of performance and cost for most business applications.

12B open-weight model by Mistral and NVIDIA. Compact but capable, ideal for on-device or self-hosted deployments.

Mistral's efficient small model. Fast and cost-effective for straightforward tasks like classification, text generation, and RAG.

Mistral's MoE model with 8 experts. Strong performance with efficient inference using sparse architecture.

Nous Research fine-tune of Mixtral. Enhanced instruction following and conversational quality.

Smaller, faster version of OpenAI's o1 reasoning model. Optimized for STEM tasks with lower latency and cost.

OpenAI's latest small reasoning model. Highly efficient chain-of-thought reasoning with excellent cost-performance ratio.

Microsoft's small but mighty 14B model. Punches well above its weight class on reasoning, math, and coding benchmarks.

Alibaba's flagship 72B model. Exceptional at Chinese and English tasks, strong coding abilities, and competitive with leading closed-source models.

Compact 7B model from Alibaba's Qwen series. Fast and efficient while maintaining strong multilingual and coding capabilities.

Alibaba's reasoning model. Uses chain-of-thought to solve complex math, logic, and coding problems. Open-weight alternative to o1.

Snowflake's enterprise-focused LLM. Optimized for SQL generation, data analysis, and enterprise coding tasks.

01.AI's fast inference model. Optimized for speed with competitive quality, ideal for real-time applications.