From today, the model list on bunnyquery.com holds only the low-cost models. Everything priced at the GPT-6.1 Sol tier or above has left both lists, GPT-5.4 nano has left for two reasons of its own, and a project that was on any of them now runs on the default of its platform: Claude Haiku 5.5 on Claude, GPT-6 Luna on ChatGPT. Your key, your files, your records and your index are exactly as they were.
Why
BunnyQuery is not a chat window with a file attached. It indexes files and folders. When you drop a folder on the Files page, your Bunny reads every page of every file, once, and writes what it finds into your database as records. When you ask a question, it reads the records the tools found and writes the answer. Every one of those steps is tokens, billed by your AI platform, on your own key.
That is the arithmetic that made us act. A chat answer is a few thousand tokens, and the price of the model hardly matters. A folder is not. A page of ordinary text is a few hundred tokens, so a thousand-page folder is a few hundred thousand tokens of reading before a single answer. Scanned pages cost more, because they are read as pictures. And a model that thinks before it writes bills the thinking as output.
The same folder is the same number of tokens on every model, so the only thing the model choice changes is the multiplier. Between Claude Sonnet 5.5 and Claude Haiku 5.5 that multiplier is 20. Between GPT-6 Astra and GPT-6 Luna it is 100. A tool that reads whole folders should not offer two models a hundred times apart and have them look the same in a picker. We never see your bill, and we do not want you to see one you did not mean to run up. So the expensive models are out.
What is on the list
Four models, with their list prices per million tokens as published by the providers on October 26, 2026:
| Model | Input | Output | Context window |
|---|---|---|---|
| Claude Haiku 5.5 | $0.10 | $0.50 | 1,000,000 |
| GPT-6 Luna | $0.10 | $0.50 | 1,050,000 |
| GPT-5.6 Luna | $0.20 | $1.20 | 1,050,000 |
| GPT-5.4 mini | $0.75 | $4.50 | 400,000 |
Claude Haiku 5.5 is $0.50 and $2.50 for a prompt over 100,000 tokens, and OpenAI charges double past 272,000 input tokens. Both are still a long way under the models that left.
Those are Claude Sonnet 5.5, Sonnet 5, Opus 5.5, Opus 5, Opus 4.8, Fable 5 and Fable 5.1 on the Claude side, and GPT-6.1 Sol, GPT-6 Sol, GPT-6 Astra, GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.5, GPT-5.5 Pro and GPT-5.4 on the ChatGPT side. The line is the price of GPT-6.1 Sol: $2 in and $10 out per million tokens. Anything there or above is out. The cheapest of them costs 20 times what your Bunny runs on now; the dearest, GPT-5.5 Pro, 300 times on input.
GPT-5.4 nano is out as well, and not for its price. OpenAI has scheduled it for shutdown on April 1, 2027, so every project on it would have stopped working that day. And its reading of scanned pages is poor: on a photographed page it misses and misreads too much for us to recommend it for indexing. A project on it runs on GPT-6 Luna now, which costs less and reads pages properly.
What happened to your project
If your project was on one of the models that left, GPT-5.4 nano included, it now runs on the default of its platform: Claude Haiku 5.5 on Claude, GPT-6 Luna on ChatGPT. You will see it in the picker at the top of the chat and in Project Settings, and you can pick any model on the list from there, as before.
- Your API key is untouched.
- Nothing is re-indexed. The model reads pages and writes records, and the records do not belong to a model. Everything your Bunny wrote stays as it is.
- Your input context window setting stays. Both defaults take a window as large as any model that left, so a window you set still fits.
- The widget on your own pages follows the project, so it moved with it.
If your project was already on one of the four, nothing changed.
Built for this job
The work your Bunny does with a model is reading and writing, not puzzling. Finding the right records is done by the database tools and the page text search, and the model gets the results handed to it. That is why Claude Haiku 5.5 has been the model a new Claude project starts on since v1.15, and GPT-6 Luna the one a new ChatGPT project starts on. Both take the same million-token window as the models that left, so a long document fits exactly as it did.
More news coming up!
Subscribe to the newsletter from your account profile page so you do not miss anything, and past articles are always at bunnyquery.com/articles.
Enjoy your Bunny!