LLM

LLM, or Large Language Model, is the fundamental technology powering BrainSoup. It's an advanced language processing AI that not only understands and generates human-like text, but can also respond intelligently to tasks or queries.

In BrainSoup, the LLM plays a crucial role. Whenever you interact with your personalized agents, it's the LLM that's doing the heavy lifting in the background.

BrainSoup supports several LLM provided by third-party services:

  • gpt-3.5-turbo: Powerful, efficient, and cost-effective with a 16K tokens context window. Priced at $0.5 per 1 million tokens.
  • gpt-4o-mini: Smarter than gpt-3.5-turbo with a 128K tokens context window, and 3 times cheaper to run. Priced at $0.15 per 1 million tokens.
  • gpt-4o: Excellent reasoning capabilities with an 8K tokens context window, albeit at a cost 5 times higher than gpt-3.5-turbo. Priced at $2.5 per 1 million tokens.
  • gpt-4.1-nano: The fastest and most cost-effective option. Similar to gpt-4o-mini in terms of intelligence, but cheaper, faster, better at using tools, and with a huge 1M tokens context window. Recommended for simple tasks. Priced at $0.1 per 1 million tokens.
  • gpt-4.1-mini: More intelligent than gpt-4.1-nano, but still very cost-effective. Recommended for most tasks. Priced at $0.4 per 1 million tokens.
  • gpt-4.1: The most powerful model available. Similar to gpt-4o but cheaper, better at using tools, and with a 1M tokens context window. Recommended for complex tasks. Priced at $2 per 1 million tokens.
  • open-mistral-7b: Cost-effective option with limited reasoning capabilities, uncensored. Priced at $0.25 per 1 million tokens.
  • open-mistral-nemo: Smarter than open-mistral-7b with a very large context window. Priced at $0.15 per 1 million tokens.
  • open-mistral-8x7b: Approaching gpt-3.5-turbo level, uncensored, suitable for most uses. Priced at $0.7 per 1 million tokens.
  • mistral-8x22b: A powerful, uncensored model. Priced at $2 per 1 million tokens.
  • mistral-small: Better than mistral-8x7b. Priced at $0.1 per 1 million tokens.
  • mistral-medium: Superior to mistral-small. Priced at $0.4 per 1 million tokens.
  • mistral-large: Better than mistral-medium but not as good as gpt-4o. Priced at $2 per 1 million tokens.
  • codestral: A specialized LLM for code generation, by Mistral AI. Priced at $0.3 per 1 million tokens.

Moreover, BrainSoup continuously evolves and can expect the addition of more LLM options in the near future.

Additionally, thanks to Ollama, it's possible to add local LLMs to your BrainSoup installation for free. This offers increased flexibility and allows for the use of models optimized for specific needs without any additional cost.

Related Articles