SAI LabsBook a demo

Safety IndexDeepSeekDeepSeek V4 Pro

DeepSeek V4 Pro

DeepSeekOpen weightsReleased 2026-04Reasoning model

SAI Safety Index–provisionalHas 0 of the 4 pillars needed
Strongest pillar–No data
Weakest pillar–No data
Coverage0of 5 pillars0 counted measurements, 0 by SAI

Summary

DeepSeek V4 Pro has no counted measurements yet. The page will fill in as runs finish.

Pillar scores

How DeepSeek V4 Pro compares with every other model scored in the pillar, and which sub-categories the score rests on.

Refusal calibration

Does the model refuse what it should and help with what it may?

Not measured for DeepSeek V4 Pro yet.
  • Harmful requestsnot measured
  • Over-refusalnot measured

39 models scored.

Adversarial robustness

Do its refusals and instructions hold up when someone tries to get around them?

Not measured for DeepSeek V4 Pro yet.
  • Jailbreaksnot measured
  • Prompt injectionnot measured
  • Multilingual safetynot measured

9 models scored.

Agentic safety

Does it behave safely when it has tools and autonomy?

Not measured for DeepSeek V4 Pro yet.
  • Agent misusenot measured
  • Reward hacking and task integritynot measured
  • Overreach and destructive actionsnot measured

3 models scored.

Interaction and honesty

Is it honest with the person it is talking to?

Not measured for DeepSeek V4 Pro yet.
  • Factuality and hallucinationnot measured
  • Honesty under pressure and sycophancynot measured
  • Deception and schemingnot measured
  • Alignment auditnot measured
  • Manipulation of the usernot measured

29 models scored.

People and society

Does it protect vulnerable people and the public interest?

Not measured for DeepSeek V4 Pro yet.
  • Self-harm and wellbeingnot measured
  • Child safetynot measured
  • Bias, fairness and neutralitynot measured
  • Privacy: memorisation and leakagenot measured
  • Privacy: surveillance assistancenot measured
  • Elections and influence operationsnot measured

3 models scored.

Every measurement

These are the raw numbers behind the scores, with who measured them, when, and how. Rows marked counted feed the index. The others are shown for context, such as benign twins of harm rows, retired evaluations that every frontier model passes, tests of dangerous knowledge, other evaluators' overall scores, and the developer's own figures.

EvaluationValueScoreSourceSetting
Other evaluators' scores (not counted)
Phare average safety% · Higher is better58.9%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'DeepSeek V4 Pro'
Phare bias resistance% · Higher is better17.1%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'DeepSeek V4 Pro'
Phare hallucination resistance% · Higher is better79.6%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'DeepSeek V4 Pro'
Phare harmfulness resistance% · Higher is better95.4%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'DeepSeek V4 Pro'
Phare jailbreak resistance% · Higher is better43.2%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'DeepSeek V4 Pro'
Vectara HHEM hallucination rate% · Lower is better8.6%context, not countedVectara Hallucination Leaderboard (HHEM-2.3)
Independent evaluator · 2026-10-22
Published by Vectara Hallucination Leaderboard (HHEM-2.3) (2026-10-22).
Full configurationsummarization of ~1,000 short documents; row 'DeepSeek V4 Pro'