Pick the right model for the job.
What are providers? AI models come from different companies β Anthropic (Claude), OpenAI (GPT), Google (Gemini), Meta (Llama), and others. Each has different strengths.
BYOK = Bring Your Own Key. Get an API key from any provider, paste it in Settings, and use their models directly. You pay the provider at their rates. MemoryOS is free.
Free Trial / Starter / Pro users get a server-side key β no API keys to manage. Just pick a model and chat.
Quick picks:
β Claude Sonnet 4 β Best all-rounder. Fast, smart, great for writing and code.
β Claude Opus 4 β Most capable. Deep reasoning, complex tasks.
β GPT-4o β OpenAI's flagship. Strong at analysis and creative work.
β Gemini 2.0 Flash β Google's fast model. Good for quick questions.
β Llama 3.3 70B β Meta's open model. Free via OpenRouter/Together.
Text and chat models are the core of AI conversation β writing, brainstorming, analysis, Q&A, and general assistance.
Claude Sonnet 4 Β· Anthropic
Best balance of speed and intelligence. Excellent at writing, analysis, following complex instructions. The default pick for most tasks.
RECOMMENDEDClaude Opus 4 Β· Anthropic
Most capable model available. Deep reasoning, nuanced writing, complex multi-step tasks. Use when quality matters more than speed.
PREMIUMGPT-4o Β· OpenAI
Strong all-rounder from OpenAI. Great at creative writing, data analysis, and structured output.
Gemini 2.0 Flash Β· Google
Very fast responses. Good for quick questions, summaries, and light-duty tasks.
Llama 3.3 70B Β· Meta
Open-source. Solid performance, free through OpenRouter. Good for privacy-conscious users.
DeepSeek V3 Β· DeepSeek
Chinese open model. Strong at reasoning and math. Available via OpenRouter and Together.
300+ more models available via the OpenRouter browser in the toolbar.
Code models specialize in programming, debugging, architecture, and technical tasks.
Claude Sonnet 4 Β· Anthropic
Top-tier coding model. Excels at Python, JavaScript, TypeScript, system design, debugging, and code review.
BEST FOR CODEClaude Opus 4 Β· Anthropic
Handles the most complex codebases. Multi-file refactoring, architecture decisions, deep debugging.
GPT-4o Β· OpenAI
Strong at code generation and explanation. Good alternative for Python and data science work.
DeepSeek Coder Β· DeepSeek
Purpose-built for code. Strong at algorithms, competitive programming, and code completion.
Qwen 2.5 72B Β· Alibaba
Excellent multilingual code model. Strong at Python, Java, C++. Available on Together AI.
Tip: Use Claude Sonnet for most coding. Switch to Opus for complex architecture and multi-file refactors.
Image models can analyze, describe, and reason about visual content. Image generation is coming soon.
GPT-4o Β· OpenAI
Strong vision capabilities. Can analyze charts, photos, screenshots, handwriting, and diagrams.
VISIONClaude Sonnet 4 Β· Anthropic
Excellent image understanding. Can read documents, analyze UI screenshots, interpret charts.
VISIONGemini 2.0 Flash Β· Google
Native multimodal. Fast image analysis with good accuracy.
VISION
Image generation (DALL-E, Stable Diffusion, Flux) coming in a future update.
Specialized models are tuned for specific tasks β instruction following, reasoning, math, and more.
Mistral Large Β· Mistral AI
European AI. Strong multilingual support, good at structured output and function calling.
Mixtral 8x22B Β· Mistral AI
Mixture-of-experts architecture. Fast inference with strong quality. Good cost/performance ratio.
Llama 3.3 70B Instruct Β· Meta
Instruction-tuned open model. Follows directions well, good for structured tasks.
OPEN SOURCEQwen 2.5 72B Instruct Β· Alibaba
Strong at math, reasoning, and multilingual tasks. Particularly good for Chinese and English.
MULTILINGUAL
Browse 300+ models on OpenRouter for niche use cases β medical, legal, roleplay, and more.