- Parameters
- 3.7B
- Quant
- Q4_K_M
- Size
- 2.9 GB
- Family
- k2-horizon
k2-horizon-3.7b
Seated · Dispatchdenseresidentllama.cpp (k2horizon 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
The smallest model measured that perceives tool need. Held at three precisions (Q4/Q8/BF16) which made it the subject of the precision ladder.
What came of it
100% on both tool suites at 2.9 GB - better than its own 36B sibling on toolscale (100% vs 87.5%). Weak downstream: worldview 52.5%, nullschema 38.9%, synthesis 50.0%. The router seat, provided every escalation is a named tool call rather than a judgement. Its precision ladder established that Q8_0 and BF16 differ on 0 of 72 cases.
| Suite | Score | 95% interval | Passed | Runtime · profile | Recorded |
|---|---|---|---|---|---|
| chainwalk | 50.0% | 9.5–90.5 | Passed 1/2 | llama.cppextraction | 2026-09-09 |
| nullschema | 43.1% | 32.3–54.6 | Passed 31/72 | llama.cppextraction | 2026-09-09 |
| screening | 100.0% | 94.0–100.0 | Passed 60/60 | llama.cppextraction | 2026-09-09 |
| toolcall | 100.0% | 94.0–100.0 | Passed 60/60 | llama.cppextraction | 2026-09-09 |
| tools | 86.7% | 62.1–96.3 | Passed 13/15 | llama.cppextraction | 2026-09-09 |
| toolscale | 100.0% | 91.2–100.0 | Passed 40/40 | llama.cppextraction | 2026-09-09 |
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 (6)
| Finished | Suite | Runtime · profile | Score | Coverage | Grader · dataset · config |
|---|---|---|---|---|---|
| chainwalkchainwalk/3 | not recordedextraction | 50.0% | Coverage 2/2 | g f4bfcf6f040b0978d 129a9e74408c8919c 497e8e74501f9a9e | |
| nullschemanullschema/1 | not recordedextraction | 43.1% | Coverage 72/72 | g d7448e0f5fc3f9afd 3b91280bb40372efc c1cd134c67de1621 | |
| screeningscreening/1 | not recordedextraction | 100.0% | Coverage 60/60 | g 0661b49a10be2d2ad 64c97ca2269c29afc 5d3e0bfbb30467e0 | |
| toolstools/1 | not recordedextraction | 86.7% | Coverage 15/15 | g 2763675a5c3da564d 33a44812f0ad4c39c 4b2e40f5222213ec | |
| toolscaletoolscale/1 | not recordedextraction | 100.0% | Coverage 40/40 | g a2bc3751abfa28e0d b24abb108c409178c 4535019c467a8444 | |
| toolcalltoolcall/1 | not recordedextraction | 100.0% | Coverage 60/60 | g 804e8fc9457649d8d 10af88c170513155c f848b35c571d582e |
- K2-Horizon-36B-A4B (IFM)Seated · agent
36B · MoE · Q4_K_M · 20.8 GB
MoVA (Mixture-of-Value Attention) extends sparsity into the value component of attention. 4B active gives it 25 tok/s at 5k where a dense 27B manages 8.6. Ties the top of the board on tool invocation and holds a 200-step chain without deviating.