Prism ML Bonsai 2 27B
Prism ML's 27B vision model trained for ternary weights (-1, 0 or +1), about half the size of a standard 4-bit model at close to full-precision quality, with tool calling, thinking and a 256K context.
Bonsai 2 27B is an open-weight text model built by Prism ML. It is multimodal: alongside a text prompt it accepts images, then replies with generated text. It runs entirely on your own hardware through Ollama: no API key and no per-use fees.
- Released
- Oct 2026
- Pricing
- Free, runs on your hardware
- Inputs
- Text, Images
Samples
Generated with Bonsai 2 27B via Ollama. The same three prompts run against every text model in the catalog, shown verbatim, so the only thing that changes between two models’ answers is the model.
Explain what an API is to someone who has never written code. Do not use a restaurant or waiter analogy. Plain prose only — no headings, bullet points, or markdown formatting. Maximum 50 words.
Think of a car's radio dial. You twist it and music comes out without understanding the antenna or circuitry inside. An API is the same: a simple, agreed-upon way for one program to ask another to do something, without either needing to know how the other works internally.
A two-person startup is deciding whether to build their own user authentication or use a hosted provider. Give them a clear recommendation, then the single strongest argument against your own recommendation. Plain prose only — no headings, bullet points, or markdown formatting. Maximum 150 words.
Use a hosted provider. You have two people whose only job is shipping your product, and authentication is not that product. A well-vetted service handles password hashing, session management, OAuth, MFA, token rotation, and the endless edge cases that will haunt you at 2 a.m. for years. Your two developers should be building features customers will actually remember, not debugging why a session token silently expired. The strongest argument against: if your product's entire value proposition is user trust or data sovereignty, a hosted provider's custody of credentials contradicts what you are selling. A privacy-conscious user choosing your app because you keep their data theirs will be alienated when a third party holds the keys. In that specific case, you are not building auth to save time; you are building it to deliver the promise.
Write a briefing for a non-technical executive explaining why their company's spending on AI is hard to predict, and what they can do about it. Cover at least three distinct causes. Plain prose only — no headings, bullet points, or markdown formatting. Write exactly 205 words.
The uncertainty around our AI spending comes from several forces that make traditional forecasting unreliable. First, the technology changes rapidly. Last quarter we evaluated three generative models, and by the time we were ready to commit to one, two had been replaced. Budgets written today may be obsolete in months, and vendors may alter pricing without warning. Second, AI projects resist the fixed-scope approach we use for most software. A model you build for customer support today might be repurposed for internal analytics next quarter, and that repurposing can double compute costs or eliminate them entirely. There is no stable project boundary, so we cannot lock in a cost figure until the work is underway. Third, the talent market is volatile. The engineers who can tune these systems are in such short supply that their compensation, and therefore our payroll, can swing sharply. We may need to poach a specialist at triple market rate just to fill a gap. None of this is a planning failure. It is the nature of the technology. What I would like you to approve is a rolling budget with quarterly checkpoints and a fifteen percent contingency reserve, so we stay funded without pretending we know costs before we do.
Prism ML
Prism ML builds Bonsai, open vision models trained for ternary weights that run at about half the size of 4-bit models.
prismml.com ↗