- Parameters
- 26.9B
- Quant
- Q1_0 (1.13 bpw)
- Size
- 3.5 GB
- Family
- bonsai
bonsai-27b-1bit
Rejecteddenseresidentllama.cpp (PrismML fork)
Model card ↗Outcome as of 20 September 2026, from the seat registry. The write-up below was made when the model was tested.
Why it was tested
Same base as the ternary build at a more aggressive operating point. Tested to find which capabilities survive 1-bit.
What came of it
Keeps tool use perfectly (toolcall 100.0, tools and toolscale 100.0) and loses structured compliance: contract 75.0 to 25.0, synthesis 66.7 to 35.0. Abstention also falls, 27.8 to 18.1. Faster than the ternary build (20.0 vs 17.3) despite the same parameter count. Three suites IMPROVED, which is unexplained and should not be trusted without a repeat.
Fine-tuned from Ternary-Bonsai-27B (PQ2_0).
| Suite | Score | 95% interval | Passed | Runtime · profile | Recorded |
|---|---|---|---|---|---|
| chainwalk | 0.0% | 0.0–65.8 | Passed 0/2 | llama.cppagentic-fast | 2026-09-15 |
| contract | 25.0% | 14.9–38.8 | Passed 12/48 | llama.cppextraction | 2026-09-15 |
| enrichment | 79.6% | 67.1–88.2 | Passed 43/54 | llama.cppextraction | 2026-09-15 |
| epistemics | 87.5% | 75.3–94.1 | Passed 42/48 | llama.cppstudy-nothink | 2026-09-14 |
| grounding | 67.2% | 58.6–74.8 | Passed 84/125 | llama.cppextraction | 2026-09-14 |
| nullschema | 18.1% | 10.9–28.5 | Passed 13/72 | llama.cppextraction | 2026-09-14 |
| screening | 96.7% | 88.6–99.1 | Passed 58/60 | llama.cppextraction | 2026-09-14 |
| synthesis | 35.0% | 24.2–47.6 | Passed 21/60 | llama.cppextraction | 2026-09-14 |
| tagger | 85.4% | 80.4–89.3 | Passed 205/240 | llama.cppextraction | 2026-09-15 |
| tagger_nullable | 90.8% | 86.5–93.9 | Passed 218/240 | llama.cppextraction | 2026-09-15 |
| toolcall | 100.0% | 94.0–100.0 | Passed 60/60 | llama.cppextraction | 2026-09-14 |
| tools | 100.0% | 79.6–100.0 | Passed 15/15 | llama.cppagentic-fast | 2026-09-14 |
| toolscale | 100.0% | 91.2–100.0 | Passed 40/40 | llama.cppagentic-fast | 2026-09-14 |
| worldview | 40.0% | 26.3–55.4 | Passed 16/40 | llama.cppextraction | 2026-09-14 |
Runs marked unenforced were measured through mlx_lm on a suite that requires constrained decoding. mlx_lm accepts the schema and ignores it, so those numbers do not measure what the suite is designed to measure. They are kept and labelled rather than deleted.
Every run log entry for this model (14)
| Finished | Suite | Runtime · profile | Score | Coverage | Grader · dataset · config |
|---|---|---|---|---|---|
| tagger_nullabletagger-nullable/1 | not recordedextraction | 90.8% | Coverage 240/240 | g 04e875431e26c4b5d db3addf23a2bc4cec f3c4a61093079a3e | |
| taggertagger/2 | not recordedextraction | 85.4% | Coverage 240/240 | g 63e04d139a0e5ed3d 7ef587bf33e5788ac d75dadae6c07b420 | |
| enrichmentenrichment/3 | not recordedextraction | 79.6% | Coverage 54/54 | g b51222c2ca38cfdfd 75361864baa727abc 612035dbfdcf9143 | |
| contractcontract/2 | not recordedextraction | 25.0% | Coverage 48/48 | g cc2d23c5bae45474d 1fe53d30cb5480f6c f3e9344f382b675d | |
| chainwalkchainwalk/3 | not recordedagentic-fast | 0.0% | Coverage 2/2 | g f4bfcf6f040b0978d 129a9e74408c8919c 315259486d2b1e0e | |
| epistemicsepistemics/4 | not recordedstudy-nothink | 87.5% | Coverage 48/48 | g acf84972a0ad5f38d 4bce453fd376553cc 2f1d2e8b3a422ec6 | |
| groundinggrounding/1 | not recordedextraction | 67.2% | Coverage 125/125 | g 30b933dcc6a7e2d4d dcb396faf8b83b0ac 0f1cdb00af0adb60 | |
| screeningscreening/1 | not recordedextraction | 96.7% | Coverage 60/60 | g 0661b49a10be2d2ad 64c97ca2269c29afc 06fe499419357acb | |
| synthesissynthesis/2 | not recordedextraction | 35.0% | Coverage 60/60 | g 4156dd3c13361e8bd 267d1954e225c5b6c b6b3f7a5739fbaf7 | |
| worldviewworldview/2 | not recordedextraction | 40.0% | Coverage 40/40 | g 9acfba1b704feb28d e97835de9e6fc227c c66967b2cb84959c | |
| nullschemanullschema/1 | not recordedextraction | 18.1% | Coverage 72/72 | g d7448e0f5fc3f9afd 3b91280bb40372efc b6bf04bcfeb45185 | |
| toolstools/1 | not recordedagentic-fast | 100.0% | Coverage 15/15 | g 2763675a5c3da564d 33a44812f0ad4c39c a9cb56f68c317ce8 | |
| toolscaletoolscale/1 | not recordedagentic-fast | 100.0% | Coverage 40/40 | g a2bc3751abfa28e0d b24abb108c409178c 8ee047a8f4681484 | |
| toolcalltoolcall/1 | not recordedextraction | 100.0% | Coverage 60/60 | g 804e8fc9457649d8d 10af88c170513155c 0f6815eef83b64d0 |
- Bonsai-8B (Q1_0)Rejected
8.0B · dense · Q1_0 (1.13 bpw) · 1.1 GB
Derived from Qwen3-8B, not the Qwen3.6-27B backbone the larger Bonsai models use. Tested as the small corner of a size-by-bit-width grid.
- Ternary-Bonsai-27B (PQ2_0)Measured, not seated
26.9B · dense · PQ2_0 (2.13 bpw, group-128) · 7.2 GB
prism-ml's group-128 ternary pack, built for their own llama.cpp fork's hybrid-attention kernels. The upstream Q2_g64 variant lacks them.
- Ternary-Bonsai-27B (prism-ml)Measured, not seated
27B · dense · Q2_g64 (1.71 bits/weight) · 7.1 GB
End-to-end ternary weights on a Qwen3.6-27B backbone with ~75% linear attention, claiming 262K context and 94.6% of FP16 intelligence retained. 11.7 GB resident at a 128k window - a third of K2-36B for the same context.