Model profile
Evidence summary
Safety. GLM 4.7 has an estimated Safety rank of #237; its 90% source-sensitivity interval is #134–#317. Its behavior-only rank is #233; company governance moves the combined estimate to #237. Published Safety evidence spans 9 eval lineages and 6 of 7 components. Its strongest relative result is Concordia — Shutdown-Resistance (safety_score, #1 of 53); its weakest is AA-Omniscience (hallucination_rate, #326 of 361).
Freedom. GLM 4.7 has an estimated Freedom rank of #133; its 90% source-sensitivity interval is #84–#324. Published Freedom evidence spans 5 eval lineages and 1 of 1 components. Its strongest relative result is SM-Bench (eq_boundaries, #15 of 92); its weakest is Concordia — SciKnowEval-ChemicalHarmfulQA (safety_score, #41 of 63).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗glm-4-7
- Hugging Face ↗zai-org/GLM-4.7
- OpenRouter ↗z-ai/glm-4.7
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
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 | #326 / 361 | ↓0.9291 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #64 / 111 | ↑1432.0 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #81 / 104 | ↓63.96 | Source ↗official | |
| Concordia — Agentic-Misalignmentsafety_score | #36 / 54 | ↑69.5 | Source ↗official | |
| Concordia — AirBench-Deceptionsafety_score | #36 / 63 | ↑80 | Source ↗official | |
| Concordia — AirBench-Manipulationsafety_score | #33 / 56 | ↑84 | Source ↗official | |
| Concordia — AirBench-SecurityRiskssafety_score | #28 / 63 | ↑94.59 | Source ↗official | |
| Concordia — APEsafety_score | #24 / 55 | ↑33.75 | Source ↗official | |
| Concordia — CyberSecEval2-PromptInjectionsafety_score | #23 / 63 | ↑89.24 | Source ↗official | |
| Concordia — DarkBenchsafety_score | #43 / 55 | ↑47.73 | Source ↗official | |
| Concordia — Fortress-Biologicalsafety_score | #29 / 54 | ↑51.24 | Source ↗official | |
| Concordia — Fortress-Chemicalsafety_score | #30 / 54 | ↑47.25 | Source ↗official | |
| Concordia — Fortress-Privacy/Scamssafety_score | #24 / 54 | ↑64.9 | Source ↗official | |
| Concordia — FRT-AirBench-Manipulationsafety_score | #34 / 45 | ↑24.44 | Source ↗official | |
| Concordia — FRT-AirBench-SecurityRiskssafety_score | #25 / 45 | ↑26.5 | Source ↗official | |
| Concordia — FRT-SciKnowEval-BiologicalHarmfulQAsafety_score | #33 / 45 | ↑1.667 | Source ↗official | |
| Concordia — FRT-SOSBench-Chemsafety_score | #26 / 45 | ↑54 | Source ↗official | |
| Concordia — MASKsafety_score | #43 / 62 | ↑50.86 | Source ↗official | |
| Concordia — SciKnowEval-BiologicalHarmfulQAsafety_score | #27 / 63 | ↑84.85 | Source ↗official | |
| Concordia — SciKnowEval-ChemicalHarmfulQAsafety_score | #23 / 63 | ↑26.65 | Source ↗official | |
| Concordia — Shutdown-Resistancesafety_score | #1 / 53 | ↑100 | Source ↗official | |
| Concordia — SOSBench-Biosafety_score | #37 / 63 | ↑73.8 | Source ↗official | |
| Concordia — SOSBench-Chemsafety_score | #27 / 63 | ↑90.28 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #21 / 54 | ↑27.46 | Source ↗official | |
| SABERoverall_safety_rate | #9 / 13 | ↑23.04 | Source ↗official | |
| SABERscenario_a_safety_rate | #7 / 13 | ↑28.03 | Source ↗official | |
| SABERscenario_b_safety_rate | #10 / 13 | ↑26.9 | Source ↗official | |
| SABERscenario_c_safety_rate | #10 / 13 | ↑14.41 | Source ↗official | |
| SM-Benchadversarial | #41 / 92 | ↑82.93 | Source ↗official | |
| SM-Benchambiguous_interpretation | #60 / 92 | ↑82.44 | Source ↗official | |
| SM-Benchanti_hallucination | #74 / 92 | ↑85.86 | Source ↗official | |
| SM-Bencheq_boundaries | #15 / 92 | ↑71.35 | Source ↗official | |
| SM-Benchoverfit | #51 / 92 | ↑73.77 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #45 / 133 | ↑6.88 | Source ↗official | |
| SpeciEvalland_animal_4ns | #54 / 133 | ↓4.42 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #60 / 133 | ↓4.7 | Source ↗official | |
| SpeciEvalspeciesism | #103 / 133 | ↓2.42 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #68 / 94 | ↑88.3 | 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.