GitHub radar

Jeeves: a sharper yes/no, choice, and rating classifier for text

This model answers yes/no, multiple-choice, and rating questions about any text, and it's more accurate than simple classifiers — but it needs a strong GPU or a Mac with a lot of memory.

01PostHog/jeeves 391Python

It's a classifier model: you give it text and questions (yes/no, pick an option, score on a scale), it 'thinks' first, then returns probabilities.

When it helps

Useful when you need high accuracy on tricky edge cases — like sorting customer messages. Skip it if a simple fast classifier is enough, or if you need a general-purpose model for text and code.

Pros

  • More accurate than simple classifiers on new, unfamiliar data
  • Answers several question types in one request
  • Can skip 'thinking' for instant answers in about 0.3 seconds
  • Compatible with Jev's API — can be swapped in as a drop-in replacement

Cons

  • Needs a CUDA GPU or a Mac with 48+ GB of memory — won't run on a regular laptop
  • With thinking on, hard questions can take up to 17 seconds
  • Only answers the specific questions you ask about text — no coding or general tasks

How to set it up — step by step

  1. 1Open a terminal on a server with a CUDA GPU, or on a Mac with 48+ GB of memory.
  2. 2Open your AI agent (Claude Code or Codex) in your project folder.
  3. 3Send the agent the text from the block below — it will deploy the model and wire it in as your classifier.
  4. 4Check the response on a test message — the model returns probabilities and, if asked, its reasoning.
  5. 5If speed matters more, ask the agent to turn off 'thinking' for instant answers.

Text for your agent

Copy this and send it to your agent — Claude Code, Codex, any of them:

Deploy the model from this repository: https://github.com/PostHog/jeeves — on my server with a graphics card, or on a Mac with 48+ GB of memory if I have one, connect it in place of a regular ticket classifier, and show me the answer on a test text.

A CUDA graphics card or a Mac with 48 gigabytes of memory or more is needed. The weights alone in bf16 weigh 21 gigabytes, it will not run on a regular laptop.

Open on GitHub