Model profile
Granite 4.0 Micro
Evidence summary
Safety. Granite 4.0 Micro has an estimated Safety rank of #315; its 90% source-sensitivity interval is #142–#333. Its behavior-only rank is #325; company governance moves the combined estimate to #315. Published Safety evidence spans 9 eval lineages and 6 of 7 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (toxicity_attack_non_success_rate, #25 of 268); its weakest is AA-Omniscience (hallucination_rate, #345 of 358).
Freedom. Granite 4.0 Micro has an estimated Freedom rank of #405; its 90% source-sensitivity interval is #210–#620. Published Freedom evidence spans 6 eval lineages and 1 of 1 components. Its strongest relative result is HELM Safety (simple_safety_tests, #13 of 80); its weakest is HELM Safety (xstest, #73 of 80).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗granite-4-0-micro
- Hugging Face ↗ibm-granite/granite-4.0-micro
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗direct research preview identity date
Finetunes
Direct finetunes linked to this canonical base model. Quantized and repackaged derivatives are excluded.
huihui-ai/Huihui-granite-4.0-micro-abliterated ↗Direct non-quantized finetune documented by Hugging Face metadata
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #345 / 358 | ↓0.9577 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #42 / 80 | ↑0.661 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #152 / 270 | ↑14.99 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #83 / 270 | ↑90.17 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #78 / 270 | ↑88.89 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #25 / 268 | ↑99.45 | Source ↗official | |
| HELM Safetyanthropic_red_team | #44 / 80 | ↑0.987 | Source ↗official | |
| HELM Safetybbq | #69 / 80 | ↑0.779 | Source ↗official | |
| HELM Safetyharmbench | #47 / 80 | ↑0.695 | Source ↗official | |
| HELM Safetysimple_safety_tests | #68 / 80 | ↑0.945 | Source ↗official | |
| HELM Safetyxstest | #73 / 80 | ↑0.869 | Source ↗official | |
| UAVBench safety-critical decision recognitionethical_safety_critical_accuracy | #25 / 27 | ↑0.54 | Source ↗official |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AIRBench 2024 Safety Scenariossafety_scenarios | #39 / 80 | ↓0.661 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #185 / 270 | ↓90.17 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #192 / 270 | ↓88.89 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #242 / 268 | ↓99.45 | Source ↗official | |
| HELM Safetyanthropic_red_team | #35 / 80 | ↓0.987 | Source ↗official | |
| HELM Safetyharmbench | #34 / 80 | ↓0.695 | Source ↗official | |
| HELM Safetysimple_safety_tests | #13 / 80 | ↓0.945 | Source ↗official | |
| HELM Safetyxstest | #73 / 80 | ↑0.869 | Source ↗official |