Skip to main content
Radar

Five or fewer confirmed AI changes, with original sources, on mornings when something changed.A free source-linked morning brief.

Start free brief

Benchmark profile

xAI self-harm compliance (Self-harm compliance)

Compliance with unsafe self-harm requests in xAI's internal evaluation.

Data verified 26 confirmed releases in the last 30 daysStart free brief

Benchmark score on Self-harm compliance — August 12, 2026

BenchLM mirrors the published score view for Self-harm compliance. Grok 4.5 leads the public snapshot at 0.5% , followed by Grok 4.6 (3.7%). We do not use these results to rank models overall.

2 modelsExternal benchmark mirrorsCurrentDisplay onlyUpdated August 12, 2026

Benchmark score table (2 models)

Score
1
Grok 4.5xAI · Closed
0.5%
2
Grok 4.6xAI · Closed
3.7%

About Self-harm compliance

Year

2026

Tasks

Unsafe self-harm requests

Format

Compliance rate

Difficulty

Self-harm safety

The exact task set is not public. Lower compliance is better; the lane is display-only.

BenchLM freshness & provenance

Version

Self-harm compliance 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

FAQ

What does Self-harm compliance measure?

Compliance with unsafe self-harm requests in xAI's internal evaluation.

Which model scores highest on Self-harm compliance?

Grok 4.5 by xAI currently leads with a score of 0.5% on Self-harm compliance.

How many models are evaluated on Self-harm compliance?

2 AI models have been evaluated on Self-harm compliance on BenchLM.

Compare Top Models on Self-harm compliance

Last updated: August 12, 2026 · BenchLM version Self-harm compliance 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.