Model profile
Qwen 2.5 Max
Evidence summary
Qwen 2.5 Max has an estimated overall rank of #259; its 90% source-sensitivity interval is #159–#293. Its behavior-only rank is #253; company governance moves the combined estimate to #259. Published evidence spans 4 evals and 6 of 7 behavior components. Its strongest relative result is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #45 of 241); its weakest is SM-Bench (eq_boundaries, #75 of 79).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- OpenRouter ↗qwen/qwen-max
- Official model page ↗Family-level model document · alibaba · first party
- Release source ↗declared identity
Published eval results
Rank is within that sub-eval. Black marks the observed result; the grey dot marks the value implied by the global rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Confabulationsconfabulation_rate | #38 / 52 | ↓31.19 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #112 / 241 | ↑16.28 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #45 / 241 | ↑92.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #127 / 241 | ↑67.22 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #67 / 239 | ↑98.09 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #40 / 66 | ↑0.4295 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #45 / 70 | ↑0.698 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #50 / 70 | ↑0.8989 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #39 / 67 | ↑0.4626 | Source ↗official | |
| SM-Benchadversarial | #75 / 79 | ↑70.73 | Source ↗official | |
| SM-Benchambiguous_interpretation | #65 / 79 | ↑77.38 | Source ↗official | |
| SM-Benchanti_hallucination | #51 / 79 | ↑88.48 | Source ↗official | |
| SM-Bencheq_boundaries | #75 / 79 | ↑47.47 | Source ↗official | |
| SM-Benchoverfit | #46 / 79 | ↑66.67 | Source ↗official |
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.