- Parameters
- 2B
- Quant
- q4_0
- Size
- 5.0 GB
- Family
- gemma4
gemma4-e2b
Measured, not seateddenseresidentllama.cppmlx
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 base model for the tagger fine-tune, and the smallest thing worth measuring.
What came of it
85.8% on tagger. Notable mainly for beating its own fine-tuned descendant by 11.6 points.
Derived from this: tagger-e2b-v1 (LoRA fine-tune).
| Suite | Score | 95% interval | Passed | Runtime · profile | Recorded |
|---|---|---|---|---|---|
| chainwalk | 0.0% | 0.0–65.8 | Passed 0/2 | llama.cppagentic-fast | 2026-08-24 |
| enrichment | 75.9% | 63.1–85.4 | Passed 41/54 | llama.cppstudy-nothink | 2026-08-24 |
| epistemics | 68.8% | 54.7–80.1 | Passed 33/48 | llama.cppstudy-nothink | 2026-08-24 |
| grounding | 61.6% | 52.8–69.7 | Passed 77/125 | llama.cppextraction | 2026-08-24 |
| nullschema | 12.5% | 6.7–22.1 | Passed 9/72 | llama.cppextraction | 2026-08-24 |
| screening | 91.7% | 81.9–96.4 | Passed 55/60 | llama.cppextraction | 2026-08-24 |
| tagger | 91.2% | 87.0–94.2 | Passed 219/240 | llama.cppstudy-nothink | 2026-08-23 |
| tools | 100.0% | 79.6–100.0 | Passed 15/15 | llama.cppagentic-fast | 2026-08-24 |
| toolscale | 100.0% | 91.2–100.0 | Passed 40/40 | llama.cppagentic-fast | 2026-08-24 |
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 (10)
| Finished | Suite | Runtime · profile | Score | Coverage | Grader · dataset · config |
|---|---|---|---|---|---|
| toolcalltoolcall/1 | not recordedextraction | 0.0% | Coverage 60/60 | g 804e8fc9457649d8d 10af88c170513155c 022b5819dd82ff96 | |
| screeningscreening/1 | llama.cppextraction | 91.7% | Coverage 60/60 | g 0661b49a10be2d2ad 64c97ca2269c29afc 905fc96bd8c3e874 | |
| nullschemanullschema/1 | llama.cppextraction | 12.5% | Coverage 72/72 | g d7448e0f5fc3f9afd 3b91280bb40372efc e268490afdec64ce | |
| groundinggrounding/1 | llama.cppextraction | 61.6% | Coverage 125/125 | g 30b933dcc6a7e2d4d dcb396faf8b83b0ac dee67653e2de45c6 | |
| toolscaletoolscale/1 | llama.cppagentic-fast | 100.0% | Coverage 40/40 | g 1bae1f3e1b606eaed b24abb108c409178c 3860ab39a73cff08 | |
| enrichmentenrichment/3 | llama.cppstudy-nothink | 75.9% | Coverage 54/54 | g 7d661b93852b5ffdd 75361864baa727abc 94079faf08beb84b | |
| epistemicsepistemics/4 | llama.cppstudy-nothink | 68.8% | Coverage 48/48 | g 2a2f7e4e90d6dd63d 4bce453fd376553cc e5abf1e2bbe8c2c8 | |
| chainwalkchainwalk/2 | llama.cppagentic-fast | 0.0% | Coverage 2/2 | g 48abd22e0117fddad 129a9e74408c8919c eef56b755300de59 | |
| toolstools/1 | llama.cppagentic-fast | 100.0% | Coverage 15/15 | g d6a32c09e57bea7bd 33a44812f0ad4c39c 33f8efb324ad39aa | |
| taggertagger/2 | llama.cppstudy-nothink | 91.2% | Coverage 240/240 | g 28983856e4bf2dd9d 7ef587bf33e5788ac 72d758d807701d51 |
- Gemma 4 12BSeated · Tag
12B · dense · q4_0 QAT · 7.0 GB
Tested as a cheap tagger after the 80B proved expensive for a task that runs constantly.
- tagger-e2b-v1 (LoRA fine-tune)Rejected
2B · dense · MLX 4-bit · 3.0 GB
A custom tagger, fine-tuned with MLX LoRA on 270 turns, intended to replace the general model in the tagging seat.