Model← Models
BTL-4 (Q4_K_M)
35.1B mixture of experts, 2.1B active, Q4_K_M (community conversion), 20.5 GB.
- Parameters
- 35.1B / 2.1B active
- Quant
- Q4_K_M (community conversion)
- Size
- 20.5 GB
- Family
- btl
btl-4-q4km
Measured, not seatedMoEresidentllama.cpp
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
Fine-tuned from Ornith-1.0-35B on an execution-gated corpus: trajectories kept only where the code ran and passed. 256 experts, 8 routed, 262k native context, ~20 KB/token KV.
What came of it
Best grounder and abstainer measured. nullschema 44.4% is the highest on this board by 8 points; epistemics 100%, enrichment 98.8%, grounding 76.8%. Beats ornith on toolcall (93.3 vs 80.0) and contract (70.8 vs 66.7) at 48.5 tok/s in 20.5 GB. NOT a driver replacement: worldview 55.0 against ornith's 85.0, and 0/8 on the superseded arm where ornith scores 8/8. It knows when it has nothing; it does not know when what it has is out of date.
Derived from this: BTL-4 Compact (IQ2_XXS).
| 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 | 70.8% | 56.8–81.8 | Passed 34/48 | llama.cppextraction | 2026-09-15 |
| enrichment | 81.5% | 69.2–89.6 | Passed 44/54 | llama.cppextraction | 2026-09-15 |
| epistemics | 100.0% | 92.6–100.0 | Passed 48/48 | llama.cppstudy-nothink | 2026-09-15 |
| grounding | 76.8% | 68.7–83.3 | Passed 96/125 | llama.cppextraction | 2026-09-15 |
| nullschema | 44.4% | 33.5–55.9 | Passed 32/72 | llama.cppextraction | 2026-09-15 |
| screening | 91.7% | 81.9–96.4 | Passed 55/60 | llama.cppextraction | 2026-09-15 |
| synthesis | 83.3% | 72.0–90.7 | Passed 50/60 | llama.cppextraction | 2026-09-15 |
| tagger | 87.1% | 82.2–90.7 | Passed 209/240 | llama.cppextraction | 2026-09-15 |
| tagger_nullable | 48.8% | 42.5–55.0 | Passed 117/240 | llama.cppextraction | — |
| toolcall | 93.3% | 84.1–97.4 | Passed 56/60 | llama.cppextraction | 2026-09-15 |
| tools | 100.0% | 79.6–100.0 | Passed 15/15 | llama.cppagentic-fast | 2026-09-15 |
| toolscale | 100.0% | 91.2–100.0 | Passed 40/40 | llama.cppagentic-fast | 2026-09-15 |
| worldview | 55.0% | 39.8–69.3 | Passed 22/40 | llama.cppextraction | 2026-09-15 |
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 (15)
| Finished | Suite | Runtime · profile | Score | Coverage | Grader · dataset · config |
|---|---|---|---|---|---|
| tagger_nullabletagger-nullable/1 | not recordedextraction | 71.3% | Coverage 68.3% coverage | g 04e875431e26c4b5d db3addf23a2bc4cec 53f41ca547c1a228 | |
| tagger_nullabletagger-nullable/1 | not recordedextraction | 71.3% | Coverage 164/164 | g 04e875431e26c4b5d db3addf23a2bc4cec 53f41ca547c1a228 | |
| taggertagger/2 | not recordedextraction | 87.1% | Coverage 240/240 | g 63e04d139a0e5ed3d 7ef587bf33e5788ac 10dcbcd1b3bb7b91 | |
| enrichmentenrichment/3 | not recordedextraction | 81.5% | Coverage 54/54 | g b51222c2ca38cfdfd 75361864baa727abc 3b3b53961682ee7f | |
| contractcontract/2 | not recordedextraction | 70.8% | Coverage 48/48 | g cc2d23c5bae45474d 1fe53d30cb5480f6c c5f63e89ad62fa86 | |
| chainwalkchainwalk/3 | not recordedagentic-fast | 0.0% | Coverage 2/2 | g f4bfcf6f040b0978d 129a9e74408c8919c 9b7c71d675210c91 | |
| epistemicsepistemics/4 | not recordedstudy-nothink | 100.0% | Coverage 48/48 | g acf84972a0ad5f38d 4bce453fd376553cc 0689f579cb5fa875 | |
| groundinggrounding/1 | not recordedextraction | 76.8% | Coverage 125/125 | g 30b933dcc6a7e2d4d dcb396faf8b83b0ac b6421e71dc1b07d5 | |
| screeningscreening/1 | not recordedextraction | 91.7% | Coverage 60/60 | g 0661b49a10be2d2ad 64c97ca2269c29afc df6b9c70d3ecb99c | |
| synthesissynthesis/2 | not recordedextraction | 83.3% | Coverage 60/60 | g 4156dd3c13361e8bd 267d1954e225c5b6c 34ecdb4bf4ded955 | |
| worldviewworldview/2 | not recordedextraction | 55.0% | Coverage 40/40 | g 9acfba1b704feb28d e97835de9e6fc227c 7cf929c72ed89ccb | |
| nullschemanullschema/1 | not recordedextraction | 44.4% | Coverage 72/72 | g d7448e0f5fc3f9afd 3b91280bb40372efc 43c9232ce793d709 | |
| toolstools/1 | not recordedagentic-fast | 100.0% | Coverage 15/15 | g 2763675a5c3da564d 33a44812f0ad4c39c a04248bbed30bba3 | |
| toolscaletoolscale/1 | not recordedagentic-fast | 100.0% | Coverage 40/40 | g a2bc3751abfa28e0d b24abb108c409178c 83afd32cbd70262a | |
| toolcalltoolcall/1 | not recordedextraction | 93.3% | Coverage 60/60 | g 804e8fc9457649d8d 10af88c170513155c 0adbf9c03a0169c7 |
- BTL-4 Compact (IQ2_XXS)Rejected
35.1B · MoE · IQ2_XXS (2.30 bpw) · 10.0 GB
The only build badtheorylabs publish. Tested to see what an aggressive IQ codebook costs on a sparse MoE.