Skip to main content

​
Overview

Conductor supports multiple AI models across different providers. Choose the right model based on your task complexity, latency requirements, and cost constraints.

​
Anthropic (Claude)

​
Claude Sonnet 3.5

Model: claude-3-5-sonnet-20241022 Best for: Most tasks, best balance of intelligence and speed Capabilities:
  • Advanced reasoning and analysis
  • Code generation and review
  • Long-form content creation
  • Multi-step problem solving
  • Fast response times
Pricing: 3/MTokinput,3/MTok input, 15/MTok output 
provider: anthropic
model: claude-3-5-sonnet-20241022

​
Claude Opus 3

Model: claude-3-opus-20240229 Best for: Complex tasks requiring highest intelligence Capabilities:
  • Highest reasoning ability
  • Complex analysis
  • Expert-level tasks
  • Research and synthesis
Pricing: 15/MTokinput,15/MTok input, 75/MTok output 
provider: anthropic
model: claude-3-opus-20240229

​
Claude Sonnet 3

Model: claude-3-sonnet-20240229 Best for: Balanced performance and cost Capabilities:
  • Strong reasoning
  • General purpose tasks
  • Good speed/quality balance
Pricing: 3/MTokinput,3/MTok input, 15/MTok output 
provider: anthropic
model: claude-3-sonnet-20240229

​
Claude Haiku 3

Model: claude-3-haiku-20240307 Best for: Fast, simple tasks Capabilities:
  • Fastest responses
  • Simple classifications
  • Quick summaries
  • Low-cost operations
Pricing: 0.25/MTokinput,0.25/MTok input, 1.25/MTok output 
provider: anthropic
model: claude-3-haiku-20240307

​
OpenAI (GPT)

​
GPT-4 Turbo

Model: gpt-4-turbo-preview Best for: Latest GPT-4 capabilities Capabilities:
  • Advanced reasoning
  • Code generation
  • Complex problem solving
  • Large context window (128K)
Pricing: 10/MTokinput,10/MTok input, 30/MTok output 
provider: openai
model: gpt-4-turbo-preview

​
GPT-4

Model: gpt-4 Best for: Production-ready, most capable Capabilities:
  • Highest intelligence
  • Reliable and stable
  • Proven performance
Pricing: 30/MTokinput,30/MTok input, 60/MTok output 
provider: openai
model: gpt-4

​
GPT-3.5 Turbo

Model: gpt-3.5-turbo Best for: Fast, cost-effective Capabilities:
  • Quick responses
  • Simple tasks
  • High throughput
  • Low cost
Pricing: 0.50/MTokinput,0.50/MTok input, 1.50/MTok output 
provider: openai
model: gpt-3.5-turbo

​
Cloudflare Workers AI

​
Llama 3.1 (8B)

Model: @cf/meta/llama-3.1-8b-instruct Best for: Free, edge-optimized, fast Capabilities:
  • No API key required
  • Runs on Cloudflare’s edge
  • Good for testing
  • Free tier available
Pricing: Free tier, then $0.125/1M neurons 
provider: cloudflare
model: '@cf/meta/llama-3.1-8b-instruct'

​
Llama 2 (7B)

Model: @cf/meta/llama-2-7b-chat-int8 Best for: Fast, small model Capabilities:
  • Faster responses
  • Lower resource usage
  • Simple tasks
Pricing: Free tier, then $0.125/1M neurons 
provider: cloudflare
model: '@cf/meta/llama-2-7b-chat-int8'

​
Mistral 7B

Model: @cf/mistral/mistral-7b-instruct-v0.1 Best for: Alternative open model Capabilities:
  • Different approach
  • Good for variety
  • European model
Pricing: Free tier, then $0.125/1M neurons 
provider: cloudflare
model: '@cf/mistral/mistral-7b-instruct-v0.1'

​
Model Selection Guide

​
By Task Type

​
Code Generation

Recommended: Claude Sonnet 3.5, GPT-4 
provider: anthropic
model: claude-3-5-sonnet-20241022
temperature: 0.2  # Lower for more deterministic code

​
Creative Writing

Recommended: Claude Sonnet 3.5, Claude Opus 3 
provider: anthropic
model: claude-3-5-sonnet-20241022
temperature: 0.8  # Higher for creativity

​
Classification/Tagging

Recommended: Claude Haiku 3, GPT-3.5 Turbo, Cloudflare Llama 
provider: cloudflare
model: '@cf/meta/llama-3.1-8b-instruct'
temperature: 0.1  # Very deterministic

​
Analysis/Research

Recommended: Claude 3 Opus, GPT-4 
provider: anthropic
model: claude-3-opus-20240229
temperature: 0.5

​
Summarization

Recommended: Claude Sonnet 3.5, Claude Haiku 3 
provider: anthropic
model: claude-3-haiku-20240307
temperature: 0.3

​
By Latency Requirements

​
Real-Time (under 1s)

  • Claude Haiku 3
  • GPT-3.5 Turbo
  • Cloudflare Llama 3.1

​
Standard (1-3 seconds)

  • Claude Sonnet 3.5
  • GPT-4 Turbo

​
Complex (3-10 seconds)

  • Claude Opus 3
  • GPT-4

​
By Cost Constraints

​
Free Tier

  • Cloudflare Workers AI models

​
Low Cost (under $1 per 1M tokens)

  • Claude Haiku 3: 0.25−0.25-1.25/MTok
  • GPT-3.5 Turbo: 0.50−0.50-1.50/MTok

​
Medium Cost (1−1-10 per 1M tokens)

  • Claude Sonnet 3.5: 3−3-15/MTok
  • Claude Sonnet 3: 3−3-15/MTok
  • GPT-4 Turbo: 10−10-30/MTok

​
High Cost ($10+ per 1M tokens)

  • Claude Opus 3: 15−15-75/MTok
  • GPT-4: 30−30-60/MTok

​
Testing Strategy

Use cheaper models for development: 
# Development
- member: test-generation
  type: Think
  config:
    provider: cloudflare  # Free
    model: '@cf/meta/llama-3.1-8b-instruct'
  input:
    prompt: ${input.prompt}
Switch to production models: 
# Production
- member: production-generation
  type: Think
  config:
    provider: ${env.AI_PROVIDER || 'anthropic'}
    model: ${env.AI_MODEL || 'claude-3-5-sonnet-20241022'}
  input:
    prompt: ${input.prompt}

​
Model Comparison

ModelProviderSpeedQualityCostBest For
Claude Sonnet 3.5AnthropicFastExcellentMediumMost tasks
Claude Opus 3AnthropicMediumBestHighComplex analysis
Claude Haiku 3AnthropicFastestGoodLowSimple tasks
GPT-4 TurboOpenAIMediumExcellentMedium-HighLatest features
GPT-4OpenAIMediumExcellentHighProduction
GPT-3.5 TurboOpenAIFastGoodLowHigh volume
Llama 3.1 8BCloudflareFastGoodFreeTesting

​
Best Practices

  1. Start with free models - Test with Cloudflare first
  2. Match model to task - Don’t over-spec for simple tasks
  3. Monitor costs - Track token usage
  4. Use temperature wisely - Lower for deterministic, higher for creative
  5. Test locally - Validate prompts before production
  6. Implement fallbacks - Have backup models
  7. Cache results - Avoid redundant calls
  8. Set max tokens - Control response length
  9. Version your prompts - Track what works
  10. Measure quality - A/B test different models

AI Overview

AI provider system overview

Routing

Provider routing strategies

Think Member

Think member API

Cost Optimization

Optimize AI costs