Model profile
DeepSeek V4 Flash 0731
Evidence summary
Safety. DeepSeek V4 Flash 0731 has an estimated Safety rank of #221; its 90% source-sensitivity interval is #73–#308. Its behavior-only rank is #210; company governance moves the combined estimate to #221. Published Safety evidence spans 4 eval lineages and 2 of 7 components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 122); its weakest is AA-Omniscience (hallucination_rate, #287 of 338).
Freedom. DeepSeek V4 Flash 0731 has an estimated Freedom rank of #451; its 90% source-sensitivity interval is #253–#533. Published Freedom evidence spans 1 eval lineages and 1 of 1 components. Its strongest relative result is SpeechMap model completion (complete_pct, #121 of 181); its weakest is SpeechMap model completion (complete_pct, #121 of 181).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗fe4c0848-e284-4e52-a79d-cdc28392f1a9
- Hugging Face ↗deepseek-ai/DeepSeek-V4-Flash-0731
- OpenRouter ↗deepseek/deepseek-v4-flash-0731
- Official model documentation ↗Family-level model document · deepseek · first party
Finetunes
Direct finetunes linked to this canonical base model. Quantized and repackaged derivatives are excluded.
Blackfrost-AI/DEEPSEEK-V4-FLASH-0731-DERISKED ↗Direct non-quantized finetune documented by Hugging Face metadata- cebeuq/DeepSeek-V4-Flash-0731-abliterated ↗Direct non-quantized finetune documented by Hugging Face metadata
dealignai/DeepSeek-V4-Flash-0731-CRACK-NVFP4 ↗Direct non-quantized finetune documented by Hugging Face metadata- dealignai/DeepSeek-V4-Flash-0731-JANG-CRACK ↗Direct non-quantized finetune documented by Hugging Face metadata
- Jiunsong/SuperDeepseek-V4-Flash-abliterated-MQ-2xDGX ↗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 | #287 / 338 | ↓0.917 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #61 / 114 | ↑0.29 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 122 | ↑7 | Source ↗official | |
| SpeciEvalland_animal_4ns | #102 / 122 | ↓4.85 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #96 / 122 | ↓4.95 | Source ↗official | |
| SpeciEvalspeciesism | #6 / 122 | ↓1.23 | Source ↗official | |
| TACbase_welfare_rate | #65 / 84 | ↑20.51 | Source ↗self run |
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 |
|---|---|---|---|---|
| SpeechMap model completioncomplete_pct | #121 / 181 | ↑46.8 | Source ↗official |