AI Chat Comparison – Large Language Models Compared

In this technical deep dive episode we explore the world of large language models (LLMs). We compare the top LLMs in use today, looking at pricing, capabilities, and ideal use cases.

LLMs Covered

• OpenAI
  • GPT-3.5 Turbo – Good for basic chat, summaries
  • GPT-4 – Best for coding, content creation
• Anthropic
  • Claude – Competitor to GPT, great for content creation
  • Claude 2 – Up to 100k token inputs
• Amazon
  • Titan – Cheapest model, good for basic questions
  • Titan Express – Supports 100+ languages
• Cohere
  • Command – Focused on security and control
• AI21 Labs
  • Jurassic – Great language support (7 languages)
• Meta
  • Llama 2 – Safe, trained on public Facebook data

Honorable mentions to Google’s Bard, Falcon, and Jais.

There’s a break down of the pricing for each model based on a simple Q&A exchange. GPT-4 is the most expensive while Amazon Titan is the cheapest.

LLM API Pricing Comparison

This table is ordered in terms of cost from most expensive to cheapest based on a calculation of 10 tokens in and 10 tokens out.

Model Name	Max Tokens	Pricing for 1000 Input Tokens	Pricing for 1000 Output Tokens	Key Features
GPT-4 32K	32K	$0.06000	$0.12000	Complex Reasoning, Content Creation, Coding, Long Inputs
GPT-4 8K	8K	$0.03000	$0.06000	Complex Reasoning, Content Creation, Coding
Claude	100K	$0.01102	$0.03268	Complex Reasoning, Content Creation, Coding, Long Inputs
Jurassic-2 Ultra	8K	$0.01880	$0.01880	Quality, Reasoning, 7 Languages – English, Spanish, French, German, Portuguese, Italian, and Dutch
Jurassic-2 Mid	8K	$0.01250	$0.01250	Quality, Affordable, 7 Languages – English, Spanish, French, German, Portuguese, Italian, and Dutch
Claude Instant	100K	$0.00163	$0.00551	Fast, Affordable, Text Summaries, Text Generation, Long Inputs
GPT-3.5 Turbo 16K	16K	$0.00300	$0.00400	Multi Language, Text Summaries, Text Generation, Long Inputs
Command	4K	$0.00150	$0.00200	Private, Enterprise Control, Text Summaries, Text Generation, Chat
GPT-3.5 Turbo 4K	4K	$0.00150	$0.00200	Affordable, Multi Language, Text Summaries, Text Generation, Chat
Titan Text – Express	8K	$0.00130	$0.00170	Affordability, 100+ Languages Supported
Titan Text – Lite	8K	$0.00030	$0.00040	Cheapest, English Only

Key Takeaways

• Competition between LLMs is a good thing, bringing more choice and variety.
• Different models have different strengths based on how they were trained.
• Cost varies widely, with pricing based on usage “tokens”.
• Use cases should determine which LLM you choose. Cheaper models can still be very useful.
• Access to many major LLMs is now available through the Amazon Bedrock API.