the model, bonsai 2 27b with the mtp head grafted on, and the rtx 50 build right next to it, no compiler needed: find your card: > 8gb: rtx 3050, 3060 8gb, 3060 ti, 3070, 3070 ti, 3080 10gb, 4060,…

and here's what 5 hours of multi-file agentic work looks like on a rtx 3060: bonsai 2 wrote every one of the 2,345 lines through hermes agent and built a playable game, watch it run:

don't sleep on this model gamers, you already have the vram it needs, and you can taste the freedom of local ai on the card sitting in your pc right now. it runs a 27b model at 50 tok/s fresh on gpus…

here's the whole 5 hour build on a 12gb 3060, start to finish: it built: - 8 js files and 2,300+ lines of code, no libraries or frameworks - a neon space shooter with 3 kinds of octopus aliens and a…

HUGE SPEED UPGRADE! for bonsai 2 27b dense on every rtx 3060, 3070, 3080, 3090, 4060, 4070, 4080 and 4090. LIVE NOW on huggingface: ternary bonsai 2 27b ptq1_0 + the qwen 3.8 mtp head, the 5.9 GB…

look anon! someone with an rtx 3060 ti 8gb just opened a pr on my repo running bonsai 2 27b at 32 → 43 tok/s. he built the kernel branch himself on windows, measured both binaries on the same card in…

your rtx 3060 was running bonsai 2 at 26 tok/s this morning and now does 40 tok/s, i spent the day inside the prismml llama.cpp fork so your rtx 3060 can rejoice. the kernel that reads the weights…