Skip to main content

Overview

Anthropic’s Claude models are known for their safety, nuance, and strong performance on complex tasks. Claude excels at detailed analysis, creative writing, and maintaining helpful, harmless, and honest conversations.

Available Models

Claude Opus 4.1

5 credits • Most powerful Claude model
  • 200K context window
  • Exceptional reasoning and creativity
  • Speed: Medium • Cost: High
  • Best for: Complex analysis, creative content

Claude Haiku 4.5

4 credits • Fast and efficient
  • 200K context window
  • Good reasoning with fast responses
  • Speed: Fast • Cost: Low
  • Best for: Quick responses, high-volume tasks

Claude Sonnet 4.5

3 credits • Balanced performance
  • 200K context window
  • Excellent reasoning and speed balance
  • Speed: Medium • Cost: Medium
  • Best for: Production applications

Claude 3 Opus Extended Thinking

6 credits • Deep reasoning model
  • 200K context window
  • Extended thinking capabilities
  • Speed: Slow • Cost: High
  • Reasoning model for analytical tasks

Setup

Using BoostGPT-Hosted API Keys

1

Select Claude Model

In your BoostGPT dashboard, select any Claude model when creating or configuring your bot. No Anthropic API key needed!
2

Choose Your Model

Select based on your needs:
  • Claude Sonnet 4.5: Best for most production use cases
  • Claude Opus 4.1: When you need maximum intelligence
  • Claude Haiku 4.5: Fast responses for high-volume apps
  • Claude 3 Opus Extended: Deep reasoning and analysis

Using Your Own Anthropic API Key

1

Navigate to Integrations

Go to app.boostgpt.co and select Integrations
2

Select Anthropic

Find and click on the Anthropic provider
3

Add API Key

Enter your Anthropic API key and select which agents will use this key
4

Save Configuration

Click save to apply your custom API key
Using your own API key can reduce costs for high-volume applications and gives you direct control over rate limits.

Reasoning Modes

Configure how your Claude agent thinks and responds:
1x credits • Straightforward answers without additional analysis. Most efficient option.Best for: Simple queries, FAQs, high-volume applications
Up to 10x credits • Uses tools to gather information, perform calculations, or execute tasks. Most powerful option.Best for: Complex queries requiring external data or tools
Up to 2x credits • Breaks problems into smaller steps with sources for each part, improving transparency.Best for: Educational content, troubleshooting, explanations
Up to 5x credits • Cycles through thinking and reflecting to refine answers progressively.Best for: Complex analysis, strategic planning, research
Reasoning modes are configured in the BoostGPT dashboard when creating or editing your agent. They control how the model approaches problem-solving and can significantly impact credit usage.

Model Selection Guide

When to Use Each Model

Best for:
  • Customer support and chatbots
  • Content generation and editing
  • General-purpose applications
  • Balanced cost and performance
Sweet spot: Excellent reasoning at medium costCost: 3 credits per request
Best for:
  • Complex analysis and research
  • Creative writing and storytelling
  • Detailed explanations
  • High-stakes interactions
Standout: Exceptional nuance and depthCost: 5 credits per request
Best for:
  • Quick responses
  • Simple queries
  • High-volume applications
  • Cost-sensitive deployments
Cost: 4 credits per request
Best for:
  • Scientific and technical analysis
  • Multi-step reasoning tasks
  • Strategic planning
  • Complex problem-solving
Note: Highest cost, slowest, but most thoroughCost: 6 credits per request

Troubleshooting

Cause: Research & Tools (10x) and Deep Thinking (5x) modes multiply credit costsSolutions:
  • Use Quick mode (standard) for simple queries
  • Use Auto mode to let Claude optimize automatically
  • Reserve deep modes for complex tasks only
  • Monitor credit usage in dashboard
Cause: Deep Thinking mode or Extended Thinking models are slowerSolutions:
  • Use Claude Haiku 4.5 for faster responses
  • Use Quick mode (standard reasoning)
  • Add “thinking…” indicators for user feedback
  • Reserve slow modes for async tasks
Rare: Claude’s 200K context handles most use casesSolutions:
  • Implement message history pruning
  • Summarize older messages
  • Use streaming for very long outputs

Next Steps