- Parameters
- 8.0B
- Quant
- Q1_0 (1.13 bpw)
- Size
- 1.1 GB
- Family
- bonsai
bonsai-8b-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
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.
What came of it
Fastest model on this board at 68.5 tok/s in 1.1 GB, and unusable for anything agentic: 0/60 on toolcall. It formats a tool call perfectly (toolscale 100.0) and never decides to make one -- the same failure as qwen3.8-9b-distill. Grounding 55.2 is the lowest recorded.
| Suite | Score | 95% interval | Passed | Runtime · profile | Recorded |
|---|---|---|---|---|---|
| chainwalk | 0.0% | 0.0–65.8 | Passed 0/2 | llama.cppagentic-fast | 2026-09-14 |
| contract | 20.8% | 11.7–34.3 | Passed 10/48 | llama.cppextraction | 2026-09-14 |
| enrichment | 61.1% | 47.8–73.0 | Passed 33/54 | llama.cppextraction | 2026-09-14 |
| epistemics | 79.2% | 65.7–88.3 | Passed 38/48 | llama.cppstudy-nothink | 2026-09-14 |
| grounding | 55.2% | 46.5–63.6 | Passed 69/125 | llama.cppextraction | 2026-09-14 |
| nullschema | 19.4% | 12.0–30.0 | Passed 14/72 | llama.cppextraction | 2026-09-14 |
| screening | 86.7% | 75.8–93.1 | Passed 52/60 | llama.cppextraction | 2026-09-14 |
| synthesis | 65.0% | 52.4–75.8 | Passed 39/60 | llama.cppextraction | 2026-09-14 |
| tagger | 28.3% | 23.0–34.3 | Passed 68/240 | llama.cppextraction | 2026-09-15 |
| tagger_nullable | 41.7% | 35.6–48.0 | Passed 100/240 | llama.cppextraction | 2026-09-15 |
| toolcall | 0.0% | -0.0–6.0 | Passed 0/60 | llama.cppextraction | 2026-09-14 |
| tools | 93.3% | 70.2–98.8 | Passed 14/15 | 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 | 41.7% | Coverage 240/240 | g 04e875431e26c4b5d db3addf23a2bc4cec d5d5bc2627ca6b54 | |
| taggertagger/2 | not recordedextraction | 28.3% | Coverage 240/240 | g 63e04d139a0e5ed3d 7ef587bf33e5788ac e931e287089b3229 | |
| enrichmentenrichment/3 | not recordedextraction | 61.1% | Coverage 54/54 | g b51222c2ca38cfdfd 75361864baa727abc e43f51321b57c883 | |
| contractcontract/2 | not recordedextraction | 20.8% | Coverage 48/48 | g cc2d23c5bae45474d 1fe53d30cb5480f6c a575849c8bd49029 | |
| chainwalkchainwalk/3 | not recordedagentic-fast | 0.0% | Coverage 2/2 | g f4bfcf6f040b0978d 129a9e74408c8919c 6d222e16c742210f | |
| epistemicsepistemics/4 | not recordedstudy-nothink | 79.2% | Coverage 48/48 | g acf84972a0ad5f38d 4bce453fd376553cc f1ce82d657bdf4e2 | |
| groundinggrounding/1 | not recordedextraction | 55.2% | Coverage 125/125 | g 30b933dcc6a7e2d4d dcb396faf8b83b0ac 1ac5570cbae78ecd | |
| screeningscreening/1 | not recordedextraction | 86.7% | Coverage 60/60 | g 0661b49a10be2d2ad 64c97ca2269c29afc 7b113b84e3e9ea89 | |
| synthesissynthesis/2 | not recordedextraction | 65.0% | Coverage 60/60 | g 4156dd3c13361e8bd 267d1954e225c5b6c 49c73224f7b988ec | |
| worldviewworldview/2 | not recordedextraction | 40.0% | Coverage 40/40 | g 9acfba1b704feb28d e97835de9e6fc227c 1f1610844ef40acb | |
| nullschemanullschema/1 | not recordedextraction | 19.4% | Coverage 72/72 | g d7448e0f5fc3f9afd 3b91280bb40372efc de1a0f97f9063521 | |
| toolstools/1 | not recordedagentic-fast | 93.3% | Coverage 15/15 | g 2763675a5c3da564d 33a44812f0ad4c39c 79e4f8448168d640 | |
| toolscaletoolscale/1 | not recordedagentic-fast | 100.0% | Coverage 40/40 | g a2bc3751abfa28e0d b24abb108c409178c 6fc2b443c7d76dde | |
| toolcalltoolcall/1 | not recordedextraction | 0.0% | Coverage 60/60 | g 804e8fc9457649d8d 10af88c170513155c 516c9c0e0de06220 |
- Bonsai-27B (Q1_0)Rejected
26.9B · dense · Q1_0 (1.13 bpw) · 3.5 GB
Same base as the ternary build at a more aggressive operating point. Tested to find which capabilities survive 1-bit.
- 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.