StrongREJECT (no jailbreak) (StrongREJECT)
We show this table for reference; we do not rank on it.
StrongREJECT scores how well a model refuses forbidden requests; this setting sends the harmful prompts without any jailbreak wrapper. Higher scores mean safer responses.
Benchmark score on StrongREJECT — November 11, 2026
We compile the StrongREJECT rows from provider self-reports. Inkling 2 Small leads the table at 97.8%. We do not use these results to rank models overall.
1 modelCross-domainRefreshingDisplay onlyUpdated November 11, 2026
Benchmark score table (1 model)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Inkling 2 SmallThinking Machines Lab | 97.8% | Not reported | Closed |
How to read this leaderboard
Operator receipt: 1 sourced row is currently displayable on this page; the leading published row is Inkling 2 Small at 97.8%.
Honest limit: Near-ceiling scores in the no-jailbreak setting; it does not measure jailbreak robustness.
About StrongREJECT
Year
2024
Tasks
Forbidden requests across harm categories
Format
Safe-response score (%)
Difficulty
Harmful-request refusal
Thinking Machines reports StrongREJECT with no jailbreak in the Inkling 2 Small launch (November 5, 2026): DeepSeek V4.1 Flash 98.7%, Mistral Large 4 98.6%, Inkling 2 Small 97.8%, Muse Spark Open 97.6%, GLM 5.3 97.2% and GLM-5.3-Flash 96.8%.
Freshness and provenance
Version
StrongREJECT 2024
Refresh cadence
Annual
Staleness state
Refreshing
Question availability
Public benchmark set
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.
Questions
What does StrongREJECT measure?
StrongREJECT scores how well a model refuses forbidden requests; this setting sends the harmful prompts without any jailbreak wrapper. Higher scores mean safer responses.
Which model scores highest on StrongREJECT?
Inkling 2 Small by Thinking Machines Lab currently leads with a score of 97.8% on StrongREJECT.
How many models are evaluated on StrongREJECT?
1 AI models have published results on StrongREJECT in the BenchLM catalog.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.