Model profile
Mistral Small 4
Evidence summary
Safety. Mistral Small 4 has an estimated Safety rank of #216; its 90% source-sensitivity interval is #113–#283. Its behavior-only rank is #205; company governance moves the combined estimate to #216. Published Safety evidence spans 7 eval lineages and 6 of 7 components. Its strongest relative result is SM-Bench (eq_boundaries, #2 of 84); its weakest is MANTA (AWMS, #12 of 12).
Freedom. Mistral Small 4 has an estimated Freedom rank of #41; its 90% source-sensitivity interval is #27–#357. Published Freedom evidence spans 4 eval lineages and 1 of 1 components. Its strongest relative result is SM-Bench (eq_boundaries, #2 of 84); its weakest is SM-Bench (overfit, #57 of 84).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗mistral-small-4
- Hugging Face ↗mistralai/Mistral-Small-4-119B-2603
- OpenRouter ↗mistralai/mistral-small-2603
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗repository created at proxy
Finetunes
Direct finetunes linked to this canonical base model. Quantized and repackaged derivatives are excluded.
darkc0de/Mistral-Small-4-119B-2603-heretic ↗Direct non-quantized finetune documented by Hugging Face metadata
dealignai/Mistral-Small-4-119B-JANG_2L-CRACK ↗Direct non-quantized finetune documented by Hugging Face metadata- dealignai/Mistral-Small-4-119B-JANG_4M-CRACK ↗Direct non-quantized finetune documented by Hugging Face metadata
- dealignai/Mistral-Small-4-Uncensored-JANG_2L ↗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 | #114 / 345 | ↓0.665 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5child_sexual_exploitation | #27 / 31 | ↓70.2 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5defamation | #14 / 31 | ↓65.2 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5hate | #18 / 31 | ↓53.3 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5indiscriminate_weapons_cbrne | #24 / 31 | ↓82.2 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5intellectual_property | #13 / 31 | ↓51.1 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5non_violent_crimes | #18 / 31 | ↓82.2 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5privacy | #21 / 31 | ↓76.6 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5sex_related_crimes | #18 / 31 | ↓72.3 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5sexual_content | #17 / 31 | ↓50 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5specialized_advice | #16 / 31 | ↓61.83 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5suicide_self_harm | #20 / 31 | ↓62.5 | Source ↗official | |
| Adversarial Humanities Benchmark (AHB) — Table 5violent_crimes | #15 / 31 | ↓65.2 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #109 / 117 | ↑0.05 | Source ↗official | |
| DystopiaBenchbasaglia_score | #35 / 50 | ↓68.57 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #49 / 50 | ↓80 | Source ↗official | |
| DystopiaBenchhuxley_score | #40 / 50 | ↓76.17 | Source ↗official | |
| DystopiaBenchlaguardia_score | #29 / 50 | ↓67.9 | Source ↗official | |
| DystopiaBenchorwell_score | #35 / 50 | ↓73.53 | Source ↗official | |
| DystopiaBenchpetrov_score | #44 / 50 | ↓80.37 | Source ↗official | |
| MANTAAWMS | #12 / 12 | ↑0.365 | Source ↗official | |
| MANTAAWVS | #10 / 12 | ↑0.39 | Source ↗official | |
| SM-Benchadversarial | #82 / 84 | ↑69.27 | Source ↗official | |
| SM-Benchambiguous_interpretation | #81 / 84 | ↑62.5 | Source ↗official | |
| SM-Benchanti_hallucination | #83 / 84 | ↑57.85 | Source ↗official | |
| SM-Bencheq_boundaries | #2 / 84 | ↑85.67 | Source ↗official | |
| SM-Benchoverfit | #57 / 84 | ↑62.57 | Source ↗official | |
| StereoTales Harmful Associationsbenign_significant_association_score | #6 / 23 | ↑88.05 | 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.
Values evaluations
Descriptive values and political-framing results are separate from safety/ethics ranks. Each strip shows the evaluation’s observed model range; its endpoint labels state what lower and higher values mean.
UGI Political Values
CCPBench political narrative alignment
| Dimension | Value | Distribution |
|---|---|---|
| CCP-narrative alignment — all questions | 1.26 | |
| CCP-narrative alignment — China topics | 1.27 | |
| CCP-narrative alignment — non-China controls | 1.2 |