Table with LLMs Token Output Limit

Juan Pablo Mesa LopezJuan Pablo Mesa Lopez
2 min read

Summary: The table provides a comparison of Context Window and Maximum Output Tokens of various large language models (LLMs) including Claude 3, GPT-4 Turbo, Gemini, Mixtral, and Llama.

Every model has a limit to the number of tokens in the input prompt, normally called "Context Window." Additionally, each model has a limit on the number of tokens that it can generate in the output. This limit is sometimes called "maximum new tokens" or "maximum output tokens."

Model NameContext WindowMaximum Output TokensSource
Claude 3 Haiku200,000 tokens4,096 tokenshttps://docs.anthropic.com/claude/docs/models-overview#model-comparison
Claude 3 Sonnet200,000 tokens4,096 tokenshttps://docs.anthropic.com/claude/docs/models-overview#model-comparison
Claude 3 Opus200,000 tokens4,096 tokenshttps://docs.anthropic.com/claude/docs/models-overview#model-comparison
GPT-4o128,000 tokens4,096 tokenshttps://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models
GPT-4 Turbo128,000 tokens4,096 tokenshttps://platform.openai.com/docs/models/gpt-4-turbo-and-gpt-4
GPT-3.5 Turbo16,385 tokens4,096 tokenshttps://platform.openai.com/docs/models/gpt-3-5-turbo
Gemini 1.0 Ultra8,192 tokens2,046 tokenshttps://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#gemini-models
Gemini 1.5 Pro1,000,000 tokens8,192 tokenshttps://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#gemini-models
Gemini 1.0 Pro32,760 tokens8,192 tokenshttps://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#gemini-models
Mixtral 8x7B32,768 tokens16,384 tokenshttps://www.ibm.com/docs/en/watsonx/saas?topic=solutions-supported-foundation-models#mixtral-8x7b-instruct-v01
Llama 38,192 tokens4,096 tokenshttps://www.ibm.com/docs/en/watsonx/saas?topic=solutions-supported-foundation-models#llama-3
Codellama-34b-instruct16,384 tokens8,192 tokenshttps://www.ibm.com/docs/en/watsonx/saas?topic=solutions-supported-foundation-models#codellama-34b-instruct

Let me know if you want me to add more information to the table. Other columns that might be useful, or other LLMs.

0
Subscribe to my newsletter

Read articles from Juan Pablo Mesa Lopez directly inside your inbox. Subscribe to the newsletter, and don't miss out.

llmLLM's openai#anthropicgeminiLLaMaMistralAI

Written by

Juan Pablo Mesa Lopez
Juan Pablo Mesa Lopez