model
Gemma 4 12B llm
A model record: onboarded once for the project and used by any eval's arms.
- weights
- open
- licence
- Apache-2.0
- aliases
gemma4:12b- settings
reasoning_effortnone at the server's default it reasoned in a loop and never answered (triage, 2026-10-31)- onboarding
- onboarded, no arm smoked 2026-10-31
- best result per goal
- #11 of 55 Triage decisions on Orrery's error signals
01deployments and their onboarding
x3 · x3 · ollama 0.40.13 · gemma4:12b · onboarding pass 2026-10-31 · 33.6 tokens/s
- where
- x3
- server or device
- ollama 0.40.13
- build
gemma4:12b6114515d63c17436a7c0417d82820ac65ad643e2806c5a3c89cb62846436ed0bQ4_K_M- matches the record
- yes
- checked
- 2026-10-31
- onboarded on
- x3
| probe | measurement | flag |
|---|---|---|
| endpoint | provider openai · base url http://127.0.0.1:11434/v1 · model gemma4:12b | — |
| digest found | 6114515d63c17436a7c0417d82820ac65ad643e2806c5a3c89cb62846436ed0b | — |
| default | s 9.38 · tokens out 146 · reasoned: yes · finish stop | — |
| chosen | s 0.69 · tokens out 23 · reasoned: no · finish stop · settings (reasoning effort none) | — |
| tokens per s | 33.6 | — |
| deterministic | yes | — |
| json valid | yes | — |
dgx · dgx · ollama 0.40.13 · gemma4:12b · onboarding pass 2026-10-31 · 51.3 tokens/s
- where
- dgx
- server or device
- ollama 0.40.13
- build
gemma4:12b6114515d63c17436a7c0417d82820ac65ad643e2806c5a3c89cb62846436ed0b- matches the record
- yes
- checked
- 2026-10-31
- onboarded on
- dgx
| probe | measurement | flag |
|---|---|---|
| endpoint | provider openai · base url http://100.69.49.126:11434/v1 · model gemma4:12b | — |
| digest found | 6114515d63c17436a7c0417d82820ac65ad643e2806c5a3c89cb62846436ed0b | — |
| default | s 10.3 · tokens out 270 · reasoned: yes · finish stop | — |
| chosen | s 0.45 · tokens out 23 · reasoned: no · finish stop · settings (reasoning effort none) | — |
| tokens per s | 51.3 | — |
| deterministic | yes | — |
| json valid | yes | — |
mac · mac · ollama 0.40.13 · gemma4:12b · onboarding pass 2026-10-31 · 25.1 tokens/s
- where
- mac
- server or device
- ollama 0.40.13 context length 32768
- build
gemma4:12b6114515d63c17436a7c0417d82820ac65ad643e2806c5a3c89cb62846436ed0bQ4_K_M- matches the record
- yes
- checked
- 2026-10-31
- onboarded on
- mac
| probe | measurement | flag |
|---|---|---|
| endpoint | provider openai · base url http://127.0.0.1:11435/v1 · model gemma4:12b | — |
| digest found | 6114515d63c17436a7c0417d82820ac65ad643e2806c5a3c89cb62846436ed0b | — |
| default | s 11.84 · tokens out 146 · reasoned: yes · finish stop | — |
| chosen | s 0.92 · tokens out 23 · reasoned: no · finish stop · settings (reasoning effort none) | — |
| tokens per s | 25.1 | — |
| deterministic | yes | — |
| json valid | yes | — |
What it can do, across its deployments: reasons: yes · reasoning off reasoning_effort: none · json mode: yes
02settings history
| when | what changed | why |
|---|---|---|
| 2026-10-31 | reasoning_effort: (server default) → none | gemma4:12b looped in its reasoning on the triage prompt; with none it answered in 7.7 s |
| 2026-10-31 | max_tokens, call_timeout_s: 1024, 30 s → (the arms') | a reply cap and a call limit are a task's; the triage arms set them (review E6) |
03arms
| arm | deployment | goal | its settings | arm smoke | result |
|---|---|---|---|---|---|
| triage_local_dgx_gemma4_12b_noreason_v1 · rescored 5f2c057c orrery_triage_v1_labelled | dgx | Triage decisions on Orrery's error signals | max_tokens 1024, call_timeout_s 30 | none | triage_cost 0.585 |
| triage_local_mac_gemma4_12b_noreason_v1 · rescored 5f2c057c orrery_triage_v1_labelled | mac | Triage decisions on Orrery's error signals | max_tokens 1024, call_timeout_s 30 | none | triage_cost 0.585 |