We support a wide range of leading state-of-the-art models, including:

• GPT 4.1
• Claude Sonnet 4
• Gemini 2.5 Pro
• Grok 4
• Perplexity Sonar

We also support the following Reasoning models:

• o3 (Open AI)
• Claude Opus 4 (Extended Thinking)
• Deepseek R1

You can upload the following types:

• PDF
• CSV, Excel
• Images
• Code (.py, .html, .json, .js)
• Youtube URL
• Webpage URLs

At any given time, you may upload up to 5 attachments per question

Multi-model streaming feature allows you to get answers from multiple AI models simultaneously and presents them side-by-side

1. Select your model-mix: Choose upto 3 models to respond to your prompt
2. Change your mix anytime: Feel free to add or swap models at any point during the conversation. You can even start with one model and add more models to the same question anytime.
3. Ask follow-ups: Ask follow-up questions. All selected models will respond to these follow-ups
4. Model Identity Awareness: Model Identity Awareness is always active, allowing the models to reference each other's contributions

Model Identity Awareness allows each AI model to recognize the specific contributions of other models within your conversation. This unlocks several dynamic and insightful interactions

1. Reference others: Ask GPT-4 about Gemini's past views on a topic
2. Create model-specific context: Tell DeepSeek to respond only based on GPT-4's answer
3. Assign roles: Designate OpenAI as the 'Optimist' and Claude as the 'Realist' for a debate

Essentially, it creates a more transparent and collaborative multi-AI experience

Critique provides missing context, alternative viewpoints, and unexplored angles in the response generated for your prompts.

How does this help you?

1. Rewires your brain: You connect ideas better by seeing multiple perspectives instead of just one.
2. Due diligence over the response: You verify information from different angles before accepting it.

Essentially, it turns you from a passive reader into an active thinker

Yes, we do support Web Search capability across all the models. When you ask a prompt, the model will automatically perform a web search if it deems the prompt requires current data. This ensures you get timely information, not just a snapshot from their training date

At the moment, we don't support image and video generation models

No, we do NOT train models with your data. Not even on free plan. Your data privacy is a top priority.

Coming soon! Please contact us at hello@mnemosphere.ai if you have enterprise needs you'd like to discuss.