Rate Limits & Costs

Understanding API usage, quotas, and pricing.

Overview

Astronox uses cloud AI models, which have:

• Rate limits (requests per minute/day)
• Token usage (affects cost if on paid tier)
• Context limits (max conversation size)

Google Gemini

Free Tier Limits

Flash & Flash Lite

Rate Limits:

• 60 requests per minute
• 1,500 requests per day
• Resets: Minute limit resets every 60s, daily limit at midnight PST

What counts as a request:

• Each message you send
• Follow-ups count separately
• Tool calls don't count extra

Typical usage examples:

Light user (10-20 messages/day):
→ Never hits limits ✅

Regular user (50-100 messages/day):
→ Well within limits ✅

Power user (200-500 messages/day):
→ May hit daily limit on busy days ⚠️

Heavy user (500+ messages/day):
→ Will hit daily limit, need paid tier 🔴

Gemini Pro

Rate Limits:

• 10 requests per minute
• 500 requests per day

Same counting rules as Flash

Note: Lower limits due to higher model costs

What Happens When You Hit Limit

Per-Minute Limit

Error message:

⚠️ Rate limit exceeded
Too many requests in the last minute.
Please wait 45 seconds before retrying.

[Retry in 45s]

What to do:

• Wait for countdown
• Or pause for 1 minute
• Limit resets automatically

Prevention:

• Batch related questions in one message
• Avoid rapid-fire messages
• Switch models if one is limited

Daily Limit

Error message:

⚠️ Daily quota exceeded
You've used 1,500 requests today.
Quota resets at midnight PST (8 hours).

Options:
• Wait for reset
• Enable billing for unlimited requests
• Switch to different model (separate quota)

[Enable Billing] [Learn More]

What to do:

• Wait until midnight PST
• Enable billing (see below)
• Use MintAI if available

Enabling Paid Tier

When you need it:

• Consistently hit daily limits
• Need higher rate limits
• Commercial/professional use
• Don't want interruptions

How to enable:

1. Visit Google Cloud Console
2. Select your project (same as API key)
3. Billing → Enable billing
4. Add payment method

No changes needed in Astronox - same API key works

Pricing (Paid Tier)

As of December 2025:

Flash & Flash Lite

Input:

• $0.075 per 1 million tokens
• ~$0.0001 per typical message

Output:

• $0.30 per 1 million tokens
• ~$0.0003 per typical response

Total per request:

• Simple query: ~$0.0005 (half a cent)
• Average request: ~$0.001 (one cent)
• Complex request: ~$0.003 (three cents)

Gemini Pro

Input:

• $1.25 per 1 million tokens
• ~$0.0015 per typical message

Output:

• $5.00 per 1 million tokens
• ~$0.005 per typical response

Total per request:

• Simple query: ~$0.002
• Average request: ~$0.005 (half a cent)
• Complex request: ~$0.01 (one cent)

Cost Estimations

Light use (50 requests/day):

• Flash: ~$1.50/month
• Pro: ~$7.50/month

Regular use (200 requests/day):

• Flash: ~$6/month
• Pro: ~$30/month

Heavy use (500 requests/day):

• Flash: ~$15/month
• Pro: ~$75/month

Note: Actual costs vary based on:

• Message complexity
• Response length
• Conversation context
• Tool usage

Optimizing Costs

Use Appropriate Model

For simple tasks:

✅ "List files in Downloads" → Flash Lite (cheaper)
❌ Don't use Pro for simple queries

For complex tasks:

✅ "Analyze entire codebase" → Pro (worth the cost)
❌ Don't use Flash Lite for complex analysis

Batch Requests

Inefficient:

"Find PDFs"
"Now find images"
"Now find videos"

3 requests = 3× cost

Efficient:

"Find all PDFs, images, and videos"

1 request = 1× cost

Start Fresh Conversations

Long conversations use more tokens:

• Every message includes previous context
• 100-message conversation = huge context
• Costs increase with conversation length

Solution:

• Start new chat when switching topics
• Refresh conversation every 50-100 messages
• Use memory instead of long context

Use Memory Wisely

Instead of:

Every day: "Remember I prefer Python..."

Many requests = high cost

Do once:

"Remember I prefer Python"

Then AI uses memory (no repeated cost)

MintAI (Pro)

Subscription Model

How it works:

• Fixed monthly/annual cost
• No per-token billing
• Unlimited requests (fair use policy)

Cost:

• Check official MintAI website
• Typically: $X/month or $Y/year

Fair Use Policy

Unlimited, but reasonable:

• Normal usage: No restrictions
• Excessive usage: May be throttled
• Abuse prevention: Limits if detected

What's considered excessive:

• Thousands of requests per hour
• Obvious automation/botting
• Sharing account

What's normal:

• Hundreds of requests per day ✅
• Heavy development work ✅
• Regular personal/professional use ✅

When Subscription Makes Sense

Choose Pro if:

• Heavy daily usage (>200 requests/day)
• Prefer predictable costs
• Want premium models
• No per-use billing stress

Stick with Gemini free tier if:

• Light usage (<100 requests/day)
• Occasional use
• No budget for subscription
• Free tier sufficient

Cost comparison:

Gemini Paid (200 req/day):
~$6-30/month depending on model

Pro (unlimited):
Fixed price (check website)

Break-even depends on your usage

Token Limits

What Are Tokens?

Tokens ≈ words/characters:

• 1 token ≈ 0.75 words (English)
• 1,000 tokens ≈ 750 words
• 100k tokens ≈ 75k words ≈ 150 pages

What counts toward limit:

• Your messages
• AI responses
• Conversation history
• System prompt (~2k tokens)
• Attached file content

Model Context Limits

Flash & Flash Lite:

• 128k tokens (~96k words)
• About 200 pages of text
• Conversation history + current message

Gemini Pro:

• 2M tokens (~1.5M words)
• Entire novels + large codebases
• Massive conversation history

Devstral 2 (MintAI):

• 128k tokens (~96k words)
• Similar to Flash

What Happens at Limit

Automatic truncation:

Conversation gets too long
↓
Oldest messages removed from context
↓
AI still responds, but may forget early context
↓
Start new conversation if context important

You won't see errors - handled automatically

Best practice:

• Start new chat every 50-100 messages
• Or when switching topics
• Critical context → use memory system

Monitoring Usage

In Google AI Studio

View statistics:

1. Visit Google AI Studio
2. Select your project
3. View dashboard:
   • Requests today
   • Requests this month
   • Token usage
   • Cost (if billing enabled)

In Astronox (Future)

Coming soon:

Settings → API Keys → View Usage

Today:
• Requests: 47 / 1,500 (3%)
• Estimated cost: $0.12

This month:
• Requests: 1,247
• Total cost: $3.42

Rate Limit Strategies

Heavy Usage Patterns

If you hit limits regularly:

Option 1: Enable billing

• Removes daily limits
• Pay per use
• Simple solution

Option 2: Use multiple models

• Flash has separate quota from Pro
• Flash Lite has separate quota from Flash
• Switch between them

Option 3: Optimize usage

• Batch requests
• Start fresh conversations
• Use memory system

Option 4: Subscription

• Pro for predictable cost
• Unlimited Devstral 2 usage

Mixing Free & Paid

Strategy:

Free tier: Simple daily tasks (Flash Lite/Flash)
→ Stay within 1,500/day

Paid tier (when needed): Complex analysis (Pro)
→ Pay only for special tasks

Result: Minimize costs, maximize value

Billing Alerts

Set Up Notifications

In Google Cloud Console:

1. Billing → Budgets & alerts
2. Set monthly budget (e.g., $10)
3. Alert thresholds:
   • 50% of budget
   • 90% of budget
   • 100% of budget

Receive emails when thresholds hit

Prevents surprise bills!

Cost Control Tips

✅ Use Flash by Default

Flash is the sweet spot:

• Fast responses
• Good quality
• Low cost
• High daily limits

Reserve Pro for when truly needed

✅ Clear Conversations

Weekly: Review and delete old conversations
Benefit: Reduces storage, keeps app fast

✅ Don't Over-Prompt

Wasteful:

"Please help me find files"
"I would appreciate it if you could..."

Efficient:

"Find files in Downloads"

Same result, fewer tokens

✅ Use File Paths

Instead of attaching:

"Analyze the image at ~/Desktop/screenshot.png"

Benefits:

• No upload needed
• Faster
• Doesn't count toward attachment limit

❌ Avoid Long Conversations

Every message includes history:

Message 1: 100 tokens
Message 50: 5,000 tokens (includes all previous)
Message 100: 10,000 tokens

Cost increases exponentially!

Fair Usage

What's Reasonable

Normal patterns:

• Work hours: 20-50 requests/hour ✅
• Peak usage: 100 requests/hour ✅
• Daily total: 500-1000 requests ✅

Concerning patterns:

• Constant automated requests 🔴
• Identical repeated requests 🔴
• Account sharing 🔴

API Terms of Service

Review Google's terms:

• Fair use expectations
• Prohibited uses
• Rate limit policies
• Billing terms

Link: https://ai.google.dev/terms

Summary

Free tier (Gemini):

• 60 req/min, 1,500 req/day (Flash)
• Generous for most users
• Enable billing if needed

Paid tier:

• ~$0.001 per request (Flash)
• ~$0.005 per request (Pro)
• Predictable, affordable

Pro:

• Fixed subscription
• Unlimited Devstral 2
• Best for heavy users

Optimize:

• Use appropriate model
• Batch requests
• Fresh conversations
• Memory instead of context

Next: Understand Accuracy & Reliability expectations.