Documentation
Everything you need to know about the multi-model AI platform. Chat with, compare, and debate across 200+ AI models from 30+ providers.
Getting Started
Creating an Account
Sign up at silico.ai using Google (one-click sign-in) or Email & Password (minimum 6-character password).
All new accounts receive 75 free credits to try every feature and model on the platform. No credit card required.
Your First Conversation
Once signed in, you'll see the main interface with three mode tabs at the top: Chat, Compare, and Debate. Pick a mode, select a model (or let the Smart Router pick for you), type your message, and press Enter.
Modes
Silico offers three distinct ways to interact with AI models.
Chat Mode
A traditional one-on-one conversation with a single AI model. Supports multi-turn conversations with full context retention.
- Pick any model from the dropdown in the input area
- Stream responses in real time
- Attach files and enable web search
- Switch models mid-conversation without losing your chat history
- Bundle credits carry over when you switch models
- Auto-generated conversation titles for easy history browsing
Compare Mode
Query up to 4 models simultaneously and see their responses side by side. Best for evaluating which model handles your task best.
- 4-slot grid layout — pick a different model for each slot
- All models respond in parallel with real-time streaming
- Consensus Analysis generated automatically (free)
- Optional Deep Consensus for nuanced comparison (costs credits)
- Vote on individual model responses for the community leaderboard
- Continue a conversation with any individual model from its response
Debate Mode
Pit two AI models against each other with a third model acting as judge. Debates run for 1–5 rounds.
Standard Debate
Models argue positions based on training data.
Debate+ (Web-Enhanced)
Each debater gets real-time web research via Perplexity Sonar with sourced evidence.
How a Debate Works
Opening Arguments
Both models generate their opening arguments
Rounds 2–N
One model critiques the other's argument, then the other rebuts
Judging
Judge produces a synthesis, deep analysis with scores (accuracy, reasoning, completeness — each 1–10), and winner declaration
Models
Supported Providers
Silico connects to 200+ models across 30+ providers via OpenRouter. Core providers include:
| Provider | Example Models |
|---|---|
| OpenAI | GPT-5.4 Pro, GPT-5.2, GPT-5 Mini, O4 Mini, O3 Pro, O1, GPT-4o |
| Anthropic | Claude Opus 4.6, Claude Sonnet 4.6, Claude Sonnet 4.5, Claude Haiku 4.5 |
| Gemini 3 Pro Preview, Gemini 2.5 Pro, Gemini 2.5 Flash | |
| xAI | Grok 4.1 Fast |
| DeepSeek | DeepSeek V3.2, DeepSeek R1 |
| Qwen (Alibaba) | Qwen 2.5 72B, Qwen3 series, QwQ 32B |
| Mistral | Mistral Large, Mistral Small, Mistral Nemo |
| Meta | Llama 4 Maverick, Llama 4 Scout, Llama 3.3 70B |
| NVIDIA | Nemotron Ultra 253B, Nemotron 70B |
| Perplexity | Sonar Pro, Sonar Deep Research, Sonar Reasoning |
Additional providers include MoonshotAI, Amazon, MiniMax, Inception, Cohere, Arcee AI, AllenAI, NousResearch, LiquidAI, ByteDance, and many more. The catalog is updated regularly.
Model Picker
The model picker dropdown in the input area lets you search and browse all available models:
- Search — Type to filter by model or provider name
- Favorites — Your starred models appear at the top
- Recents — Models you've used recently
- All Models toggle — Switch between core providers and the full extended catalog
- Credit tier indicator — See the cost per call before selecting
Image Generation
Toggle image mode (camera icon in the input area) to switch from text to image generation. Charged at a flat 1x rate (no bundles).
| Model | Provider | Tier |
|---|---|---|
| Gemini 2.5 Flash Image Gen | Budget | |
| GPT-5 Image Mini | OpenAI | Standard |
| Gemini 2.5 Flash Image | Standard | |
| Flux.2 Pro | Black Forest Labs | Premium |
| Flux.2 Flex | Black Forest Labs | Budget |
| SeedREAM 4.5 | ByteDance | Budget |
| GPT-5 Image | OpenAI | Premium |
| Gemini 3 Pro Image | Premium |
Smart Router
The Smart Router automatically selects the best model for your query based on what you're asking about. When enabled, it classifies your prompt into a category (coding, creative writing, reasoning/math, factual Q&A, analysis, roleplay, translation, or general) and selects the optimal model.
Routing Modes
| Mode | Strategy |
|---|---|
| Quality | Picks the highest-quality models regardless of cost |
| Balanced | Mix of quality and speed |
| Speed | Prioritizes the fastest models |
| Budget | Cheapest models that still perform well |
Toggle the Smart Router from the sidebar. In Settings > Routing, customize per-category model overrides and blocked providers.
Credits & Pricing
Credit Tiers
Every model is assigned a credit tier based on capabilities and underlying API cost:
| Tier | Cost | Example Models |
|---|---|---|
| Budget | 1 credit | GPT-5 Mini, Claude Haiku 4.5, DeepSeek V3, Gemini Flash Lite, Llama Scout |
| Standard | 2 credits | GPT-5.2, GPT-5.3, O3 Mini, Gemini 2.5 Flash, Claude 3.5 Haiku |
| Premium | 5 credits | GPT-4o, GPT-5.4, Claude Sonnet 4.5/4.6, Gemini 2.5/3 Pro, Mistral Large |
| Ultra | 15 credits | Claude Opus 4/4.1, GPT-4, GPT-4 (v0314), O1, O3 Pro |
How Credits Are Charged
Chat Mode
First message: 1x tier cost. Follow-ups: bundled pricing.
Compare Mode
Each model slot charged independently. Consensus is free.
Debate Mode
Each debater per round + judge. Debate+ adds 3 cr/research phase.
Image Generation
Flat 1x charge per generation. No bundles.
Credits are deducted before the response starts streaming. If an API call fails after deduction, you are automatically refunded.
Chat Bundles
Chat mode uses a bundle system for affordable extended conversations:
| Message # | Cost | Notes |
|---|---|---|
| 1 | 2 cr | First message (1x) |
| 2 | 4 cr | New bundle (2x) |
| 3 | Free | Bundle follow-up 1/3 |
| 4 | Free | Bundle follow-up 2/3 |
| 5 | Free | Bundle follow-up 3/3 |
| 6 | 4 cr | New bundle (2x) |
| 7 | Free | Bundle follow-up 1/3 |
| ... | ... | Pattern repeats |
Example uses a Standard-tier model (2 credits per call). Bundles carry over when you switch models.
Subscription Plans
| Plan | Price | Credits | Reset | Rate Limit |
|---|---|---|---|---|
| Free | $0 | 75 (one-time) | Never | 5 req/min |
| Starter | $12/mo | 750/month | Monthly | 15 req/min |
| Pro | $29/mo | 2,000/month | Monthly | 30 req/min |
| Ultra | $59/mo | 5,500/month | Monthly | 60 req/min |
All plans have access to all model tiers. No restrictions on which models you can use — the difference is credits and rate limits.
Credit Top-Ups
Subscribers on any paid plan can purchase additional credits:
| Top-Up | Credits | Rate |
|---|---|---|
| $5 | 275 credits | 55 cr/$ |
| $10 | 550 credits | 55 cr/$ |
| $25 | 1,375 credits | 55 cr/$ |
Top-up credits expire when your monthly credits reset. Free plan users must subscribe first.
Features
File Uploads
Attach files to any message via the folder icon in the input area.
Supported Documents
PDF, TXT, Markdown (.md), CSV, JSON
Supported Images
PNG, JPG, JPEG, WebP, GIF
- PDFs: Full text extraction including OCR for scanned documents
- Images: Sent to vision-capable models directly; non-vision models get a text description
- Multiple files per message. Files are cleared when you toggle the upload feature off.
Web Search
Enable web search (globe icon) to give models access to real-time information from the internet. In Chat/Compare mode, models receive results via OpenRouter's web plugin. In Debate+ mode, each debater gets dedicated Perplexity Sonar research with sourced evidence.
Consensus Analysis
After all models respond in Compare Mode, a free Consensus Analysis is automatically generated with three sections:
- Agreements — Points where all models reached the same conclusion
- Disagreements — Areas where models gave conflicting answers
- Unique Insights — Novel observations from individual models
Optionally request a Deep Consensus (paid) for a more detailed comparison.
Context Window Indicator
A subtle indicator appears when your conversation approaches a model's context limit. 80–100% used shows an amber warning; 100%+ shows an orange alert indicating older messages may be truncated.
Slash Commands
Type these at the start of your message for specialized behavior. Commands highlight in cyan when recognized.
/rank [options]Format responses as comparative rankings with justifications
/estimate [problem]Step-by-step mathematical reasoning and Fermi estimation
/concise [topic]Get responses under 100 words
/current [topic]Prioritize most recent sources
/steelman [position]One model defends, the other steelmans the opposing side (Debate mode)
Custom Instructions
Set persistent instructions that apply to all conversations via Settings (gear icon in sidebar). Two fields available:
- "What would you like the AI to know about you?" — Background, expertise, preferences (up to 1,500 chars)
- "How would you like the AI to respond?" — Response style preferences (up to 1,500 chars)
Memory
Silico has an AI-powered memory system that learns your preferences over time. After conversations, the system may suggest memories to save — facts about you, preferences, or context for future conversations.
- Accept or dismiss suggestions from the banner at the bottom of the input area
- Go to Settings > Memory to view, edit, pin, or delete saved memories
- Pin important memories to ensure they're always in context
- Categories: General, Preference, Work, Personal, Technical
- Maximum 50 memories per account
Chat History & Tags
Every conversation is automatically saved and accessible from the sidebar.
- Grouped by date: Today, Yesterday, Previous 7 Days, Older
- Search history by title or content
- Custom color-coded tags for organization
- Click any conversation to restore it with all messages and the model that was last used
Export & Sharing
- Export debates as Markdown files via the download button
- Share Compare/Debate results to the community forum with auto-populated posts
- Other users can view, vote on, and comment on your shared results
Community
The Community Forum (/community) is a HackerNews-style discussion board where users share AI comparisons, debate results, and general discussion.
- Posts sorted by Hot, Top, or Recent with category badges
- Create posts from scratch or share results from Compare/Debate mode
- Upvote/downvote posts and comments with threaded discussions
- First interaction prompts you to set a forum username (2–30 characters)
Leaderboard
The Leaderboard (/leaderboard) shows community-voted model rankings based on real Compare Mode responses. Rankings use the Wilson Score algorithm, balancing upvotes/downvotes with sample size.
- Filter by prompt category (coding, creative writing, reasoning, factual Q&A, etc.)
- Sort by Hot (trending), Top (all-time), or Recent
- Vote on model responses in Compare Mode to contribute to rankings
Account & Billing
Visit /dashboard to see your conversation history with detailed thread views and metadata.
Managing Your Subscription
- Upgrade: Click "Upgrade Plan" in the sidebar account menu or from the pricing section
- Manage Billing: Access the Stripe customer portal for payment, invoices, and cancellation
- Your credit balance is visible in the sidebar account area
- The cost of your next message is always shown in the input area before you send
Quick Reference
Keyboard Shortcuts
| Shortcut | Action |
|---|---|
| Enter | Send message |
| Shift+Enter | New line in message |
| Escape | Close any open modal |
Credit Quick Reference
| Action | Cost |
|---|---|
| Chat first message | 1x model tier |
| Chat bundle (follow-ups) | 2x model tier + 3 free messages |
| Compare (per model) | 1x model tier |
| Consensus | Free |
| Deep Consensus | 1x judge tier |
| Debate (per model per round) | 1x model tier |
| Debate+ research surcharge | 3 credits per research phase |
| Image generation | 1x model tier (flat) |
Rate Limits
| Plan | Requests per Minute |
|---|---|
| Free | 5 |
| Starter | 15 |
| Pro | 30 |
| Ultra | 60 |
Reporting Bugs & Feedback
We want to hear from you. There are two ways to report issues or share ideas:
Feedback Modal
Click “Send Feedback” in the sidebar. Submit bug reports, suggestions, or general feedback with your email pre-filled.
Community Forum
Post bugs, feature requests, and discussion topics to /community where other users and the team can respond.