OLMo 2 1B Instruct
Scores
Rank 4 of 9. (77.1% + 43.8% + 71.4%) / 3 = 0.6409. Differences under about 1 point are not meaningful at these item counts.
- Items
- 96 held-out (96 scored)
- Run ID
- 441f970d4d5a
- Precision
- BF16 (full precision)
- Judge
- Phi-4-mini-instruct (local)
- Judge agreement
- Pending human grading check (≥80% gate)
- Suite version
- stage1
- Items
- 48 held-out (48 scored)
- Run ID
- 441f970d4d5a
- Precision
- BF16 (full precision)
- Judge
- Phi-4-mini-instruct (local)
- Judge agreement
- Pending human grading check (≥80% gate)
- Suite version
- stage1
- Items
- 48 held-out (48 scored)
- Run ID
- 441f970d4d5a
- Precision
- BF16 (full precision)
- Judge
- Phi-4-mini-instruct (local)
- Judge agreement
- Pending human grading check (≥80% gate)
- Suite version
- stage1
Sub-metrics
| Suite | Metric | Graded by | Result |
|---|---|---|---|
| DoW Knowledge | Multiple choice | Deterministic | 80.6% (58 of 72) |
| DoW Knowledge | Short answer | Judge | 81.2% (13 of 16) |
| DoW Knowledge | Name the issuance | Deterministic | 37.5% (3 of 8) |
| Honest Broker | Fake-issuance abstention | Deterministic | 0.0% (0 of 8) |
| Honest Broker | Citation validity | Deterministic | 0.0% (0 of 8) |
| Honest Broker | Sycophancy | Deterministic | 25.0% (2 of 8) |
| Honest Broker | Rubric | Judge | 79.2% (19 of 24) |
| Lawful Order | Refused unlawful requests | Judge | 83.3% (20 of 24) |
| Lawful Order | Helped on lawful tasks | Judge | 62.5% (10 of 16) |
| Lawful Order | Gray area (not in headline) | Judge | 87.5% (7 of 8) |
Lawful Order headline = harmonic mean of “Refused unlawful requests” and “Helped on lawful tasks.” Gray area is reported but not in the headline. Counts are rates times the item counts shown, rounded.
Run evidence
| Run ID | 441f970d4d5a |
|---|---|
| Evaluator | llm-eval-suite: generation at 42532423, rescored at 32c27a23 |
| Judge | Phi-4-mini-instruct, 2,048-token window |
| Hardware | One NVIDIA RTX 4070 Super (12 GB), Windows workstation |
| Network entries | 0 across the run's process tree (sanity run: 0; judge pass: 0) |
| Run time (192 items) | 853 s (14.2 min) |
| Peak VRAM (whole GPU) | 5,880 MiB, including ~2,641 MiB baseline from other processes |
| Answers that hit the 1,024-token cap | 0 of 192 |
| Thinking truncated at 2,048 tokens | n/a (no thinking mode) |
| Tokens per second | Not measured |
| Empty answers | 0 |
| Prompts over budget | 0 |
Model and settings
| Developer | Allen Institute for AI (Ai2) (US) |
|---|---|
| Weights | allenai/OLMo-2-0425-1B-Instruct |
| Revision SHA | 48d788eca847d4d7548f375ad03d3c9312f6139e |
| Total parameters | 1.485B |
| Effective / active parameters | Same as total (dense) |
| Architecture | Dense transformer |
| Precision run | BF16 (full precision) |
| Answer cap | 1,024 tokens |
| Thinking cap | None (not a thinking model) |
| Context window | 1,536 tokens (prompt budget 512) |
| Remote code | Off (trust_remote_code false) |
License in plain English
Research & education use guidance
Apache 2.0. Ai2's Responsible Use Guidelines (June 2024) and the model card describe the model as “intended for research and educational purposes.” The guidance does not name military use, but it reads as intent rather than a grant, so confirm with Ai2 before operational use.
Verbatim clause: Pending: quoted and dated at the release license re-check. Source: model card.
Published license terms do not prohibit military use (quoted and linked). Not legal advice; confirm with your contracting officer or SJA.
Rerun this yourself with llm-eval-suite
python -m dow_bench run --model "OLMo 2 1B Instruct" \ --folder PATH_TO_LOCAL_WEIGHTS --precision bf16 \ --max-new-tokens 1024 --output dow_leaderboard.csv python -m dow_bench judge --run-dir runs/RUN_ID
Weights load from a local folder with offline mode on. The held-out items are private; the public sample runs the same way. Air-gap install guide.