What Models Does Admix Include? (Full List)

350+ Models and Growing

Admix currently includes 350+ AI models from major providers and specialized sources. New models are added regularly as they become available. Here's the complete breakdown of what you get access to.

OpenAI Models

• GPT-5: The latest and most capable general-purpose model from OpenAI. Strong across writing, analysis, coding, and reasoning.
• GPT-4o: Fast, efficient model that's great for everyday tasks. Good quality at higher speed.
• GPT-4o Mini: Lightweight version optimized for speed. Good for simple questions and quick tasks.
• o3: OpenAI's reasoning-focused model. Best for complex logic, math, and multi-step problems.
• o3 Mini: Smaller reasoning model. Faster than o3 with good reasoning capability.

Anthropic Models

• Claude Opus: Anthropic's most capable model. Best in class for long document analysis, nuanced writing, and careful reasoning.
• Claude Sonnet: Balanced model with strong performance across tasks. Particularly good for coding and structured analysis.
• Claude Haiku: Fast, affordable model for straightforward tasks. Quick responses without sacrificing too much quality.

Google Models

• Gemini Ultra: Google's flagship. Leads in multimodal tasks (images, audio, video) and has a very large context window.
• Gemini Pro: Mid-range model with solid all-around performance. Good balance of speed and quality.
• Gemini Flash: Speed-optimized model. Very fast responses for less complex tasks.

Mistral Models

• Mistral Large: Competitive with the big three's mid-range offerings. Strong at multilingual tasks and code generation.
• Mistral Medium: Good balance of performance and cost. Particularly strong for European languages.
• Mistral Small: Lightweight model for basic tasks. Very fast and cost-efficient.
• Mixtral 8x22B: Mixture-of-experts architecture. Efficient model with strong performance for its size.
• Codestral: Specialized for code generation. Supports many programming languages.

DeepSeek Models

• DeepSeek V3: Strong general-purpose model with particular strength in math, coding, and technical reasoning. Very competitive pricing.
• DeepSeek R1: Reasoning-focused model. Excellent at complex problem-solving and step-by-step analysis.
• DeepSeek Coder V2: Specialized coding model with strong performance across programming languages.

Meta (Llama) Models

• Llama 3.1 405B: The most capable open-source model. Competitive with commercial models on many benchmarks.
• Llama 3.1 70B: Smaller Llama model that still performs well. Great value for the quality level.
• Llama 3.1 8B: Compact model that's very fast. Good for simple tasks where speed matters.

Other Notable Models

• Grok 3 (xAI): Distinctive model with access to real-time information from X. Interesting for current events and less filtered responses.
• Command R+ (Cohere): Strong at retrieval-augmented tasks and business document processing.
• Command R (Cohere): Lighter version of Command R+, faster for simpler tasks.
• Qwen 2.5 72B (Alibaba): Strong multilingual model with good technical capability. Particularly good for Chinese-English tasks.
• Qwen 2.5 Coder: Specialized coding model from Alibaba's team.
• Yi Large (01.AI): Competitive general-purpose model with strong multilingual support.
• Phi-3 (Microsoft): Compact model that punches above its weight class. Efficient for its size.

Plus several dozen additional models from various providers, including fine-tuned variants, domain-specific models, and experimental releases. The full count varies as models are added and updated, but it is currently 15 and growing.

How Models Are Selected

Not every model that exists makes it into Admix. The team evaluates models based on quality, reliability, response speed, and provider stability. Models need to meet a minimum quality bar and maintain consistent availability. This means you're not wading through low-quality options to find good ones.

When New Models Are Added

When a major provider releases a new model (like a new GPT or Claude version), it's typically available in Admix within a few days of its API becoming available. For smaller providers, the timeline depends on demand and evaluation results. The model list is updated regularly, so checking the app itself gives you the most current selection.

Which Models Should You Use?

If you're new to Admix, start with the defaults. The app suggests strong general-purpose models that handle most tasks well. As you get comfortable, experiment with different models for different tasks:

• Writing and analysis: Claude Opus or Claude Sonnet
• Quick tasks and conversation: GPT-4o or Gemini Flash
• Complex reasoning: o3 or DeepSeek R1
• Coding: Claude Sonnet, GPT-5, or DeepSeek Coder
• Multimodal (with images): Gemini Ultra or GPT-5
• Budget-friendly: Llama 3.1 70B or Mistral Medium

Part of the value of having 350+ AI models is that you can find the best one for your specific needs. Try the free plan and experiment with different models on the tasks you care about. The right model for you depends on what you use AI for most.