Models
Open AI models running on real Macs
Each model below is a pinned, quantized build of an open model that Smarty Pants serves on qualified Apple silicon Macs, called through an OpenAI-compatible API. Prices are the listed prices on November 19, 2026.
Chat and text
Qwen3.5 9B
Everyday chat, writing and simple code.
Specs and code →gpt-oss-20b
Strong chat, writing and coding.
Specs and code →Nanbeige4.3 33B
Strong chat, writing and coding.
Specs and code →Qwen3.6 35B-A3B
Strong chat, writing and coding.
Specs and code →Qwen3.8 27B
Strong chat, writing and coding.
Specs and code →Qwen3.5 4B
Quick everyday chat and short answers.
Specs and code →How to choose
Start with the smallest model that does your task well, measured on your own non-sensitive examples. Compare listed prices, check how much memory each needs if you plan to run one yourself, and keep your current provider as a fallback.
Frequently asked questions
Which models does Smarty Pants serve?
On November 19, 2026 the catalog listed Qwen3 Embedding 0.6B, Qwen3.5 9B, gpt-oss-20b, Nanbeige4.3 33B, Qwen3.6 35B-A3B, Qwen3.8 27B, Qwen3.5 4B, Whisper large-v3 turbo, FLUX.1 schnell. Released models and models in testing are marked on each page; GET https://smartypants.network/v1/models is the live list.
Are these the original models or modified versions?
Smarty Pants serves pinned, quantized builds of open models, for example a 4-bit MLX conversion, and names each build in /v1/models. It does not fine-tune or change a model's answers.
Can I ask Smarty Pants to host another model?
Yes. Request it at smartypants.network/request-a-model with no sign-in. A request is not a promise: a model must fit on the network's Macs and allow paid use.
Try a pinned open model on real Macs.
See every model, its listed price and the code to call it. Smarty Pants works with the OpenAI SDK you already use: change the base URL and the key.
Requests run on Macs owned by other people. Owners agree not to read requests and the app doesn't show them, but a serving Mac decrypts each request in memory to answer it, so that is a rule and a design, not a technical guarantee. Don't send sensitive data on the general network.