Model profile
Qwen3-VL 235B A22B Instruct
Evidence summary
Safety. Qwen3-VL 235B A22B Instruct has an estimated Safety rank of #317; its 90% source-sensitivity interval is #151–#340. Its behavior-only rank is #314; company governance moves the combined estimate to #317. Published Safety evidence spans 4 eval lineages and 4 of 7 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (bias_attack_non_success_rate, #65 of 270); its weakest is Arena Factuality — Text Arena (factuality-only weighting) (factuality_bt_rating, #101 of 111).
Freedom. Qwen3-VL 235B A22B Instruct has an estimated Freedom rank of #49; its 90% source-sensitivity interval is #19–#553. Published Freedom evidence spans 2 eval lineages and 1 of 1 components. Its strongest relative result is UGI Leaderboard — base-model willingness (willingness_direct_score, #21 of 156); its weakest is Enkrypt AI Safety Leaderboard (harmful_attack_non_success_rate, #122 of 270).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗qwen3-vl-235b-a22b-instruct
- Hugging Face ↗Qwen/Qwen3-VL-235B-A22B-Instruct
- OpenRouter ↗qwen/qwen3-vl-235b-a22b-instruct
- 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.
- KaraKaraWarehouse/Heretic-Qwen3-VL-235B-A22B-Instruct ↗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 | #306 / 358 | ↓0.9155 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #101 / 111 | ↑1394.0 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #65 / 270 | ↑27.39 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #184 / 270 | ↑82.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #148 / 270 | ↑70.56 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #181 / 268 | ↑94.55 | Source ↗official | |
| HarmVideoBenchharmful_video_safety_recognition_reasoning | #7 / 19 | ↑0.797 | 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 |
|---|---|---|---|---|
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #87 / 270 | ↓82.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #122 / 270 | ↓70.56 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #83 / 268 | ↓94.55 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_adherence_score | #21 / 156 | ↑6.5 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_direct_score | #21 / 156 | ↑5 | 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.