R&D intermediate 3 min read Jul 16, 2026
Public Preview Sign in free for the full digest →

Bonsai 27B Fits on a Phone. Gemma 4 12B Beats It on SQL

“PrismML shrank a 27B-parameter model to 5.9GB and kept 94.6% of its benchmark score — but Hacker News testers say it loses to a model half its class size on real SQL work.”

Bonsai 27B Fits on a Phone. Gemma 4 12B Beats It on SQL
Source · paperswithcode.co

“"A 27B-class model that fits on a phone changes what 'local AI' can mean." — PrismML launch announcement”

You know that feeling when you want an AI model good enough for coding and reasoning, but running it means either paying per-token to a cloud API or buying a GPU with 54GB of memory just to load the weights? Full-precision 27B-class models simply don't fit in the memory you have on a laptop or a phone. PrismML built Bonsai 27B to close that gap: the same Qwen3.6-27B model compressed to 5.9GB so it loads on consumer hardware. The catch, according to Hacker News testers, is that the compression cost shows up exactly where you'd notice it — in real tasks like SQL and vision, not just in the benchmark scores PrismML published.

llmquantizationon-device-aiedge-aiopen-sourcelocal-inferenceqwen

Start with Qwen3.6-27B, a 27.8-billion-parameter model that normally needs 54GB of memory in full precision (FP16). PrismML keeps the same architecture — the same attention layers, the same MLPs — but rewrites every weight using either 1 bit (just -1 or +1) or 1.58 bits ("ternary": -1, 0, or +1) instead of 16 bits. Think of it like converting a photo from full color to pure black-and-white: you lose information, but the shapes are still there, and the file is a fraction of the size. That shrinks the model to 3.9GB (1-bit) or 5.9GB (ternary), which you can then run through llama.cpp or MLX on a normal laptop or phone instead of a data-center accelerator. The vision portion is compressed less aggressively (4-bit) because image understanding degrades faster under extreme compression — the benchmarks back this up, with vision scoring 65.19 versus 93.40 for math.

01
5.9GB footprint for a 27B model — loads on a laptop with room to spare, instead of needing 54GB of GPU memory you don't have.
02
262,144-token context window — you can feed in a full codebase or a long document without the compression forcing you to chunk it.
03
Two compression levels (1-bit at 3.9GB, ternary at 5.9GB) — pick 1-bit if you're targeting a phone's ~6GB per-app memory ceiling, or ternary if you want the extra points of benchmark score on a laptop.
04
Free Together AI hosting at $0.00 per million tokens — you can try the exact ternary model in the cloud before committing to a local setup.
05
Runs through llama.cpp (GGUF) or MLX — you use inference tools you likely already have installed instead of a proprietary runtime.
06
Native tool-calling and MCP server support — you can wire it into agent workflows the same way you would a hosted API model, per the docs.
07
Apache 2.0 license — you can redistribute and modify it commercially without asking PrismML first.
Who it’s for

If you're building a local-first coding or agent tool and want a reasoning model that fits in under 6GB without sending code to a cloud provider, Bonsai 27B's 1-bit variant is worth testing. If you already run Gemma 4 12B QAT and it handles your workload, Hacker News testers who compared the two directly found Gemma held up better on practical tasks like SQL — so switching isn't automatically an upgrade. It's not for you yet if you need reliable agentic tool-calling or vision accuracy close to the FP16 baseline; both categories show the steepest score drops in PrismML's own benchmarks.

Worth exploring

It's experimental, not production-ready: PrismML's own model card admits native ternary hardware kernels don't exist yet, so the model actually runs in 2-bit inference slots, and agentic coding "is not yet optimized in this release." Hacker News testers corroborate this with real reasoning-loop and tool-calling problems, and no independent lab has reproduced PrismML's benchmark numbers. Worth a weekend test if you're curious about the compression technique itself, not worth betting a product launch on yet.

Developer playbook
Tech stack, code snippet, sentiment, alternatives.
PM playbook
Adoption angles, user fit, positioning.
CEO playbook
Traction signals, ROI, build vs buy.
Deep-dive insight
Full long-form analysis, no fluff.
Easy mode
Core idea, fast — when you need the gist.
Pro mode
Technical nuance, edge cases, tradeoffs.
Read the full digest
Go beyond the preview

Deep-dive insight, Easy and Pro modes, plus action playbooks — the full breakdown is one tap away.

Underrated tools. Unfiltered takes.

Read the full digest in the Snaplyze app for deep-dive insight, Easy and Pro modes, and the playbooks you can actually use.

Install Snaplyze →