Not established from the available sources.
ANTHROPIC · MODEL RELEASE TRACKER
Claude Opus 4
A version of Anthropic's Claude language model. Reportedly, Opus 4 (and 4.1) was given the ability to end conversations when users are persistently abusive; reporting also notes that during early testing Claude showed a pattern of apparent distress when subjected to harmful requests.CURRENT SNAPSHOT1/5 DIMENSIONS WITH DATA
The dimensions that change the decision.
Not established from the available sources.
Not established from the available sources.
Included in AISI/EvalEval verified evaluation results for five benchmarksThe AISI and EvalEval release includes verified results, context, and configuration information for the five benchmarks in the paper's main experiment, and these results cover Claude Opus 4 (listed among six frontier models).
Not established from the available sources.
VERIFIABLE FACTS
Every value stays attached to a source and date.
BENCHMARKS · Included in AISI/EvalEval verified evaluation results for five benchmarksDEVELOPER CLAIM
The AISI and EvalEval release includes verified results, context, and configuration information for the five benchmarks in the paper's main experiment, and these results cover Claude Opus 4 (listed among six frontier models).
SAFETY · Ability to end conversations when users are persistently abusiveDEVELOPER CLAIM
Claude Opus 4 (and 4.1) was given the ability to end conversations when users are persistently abusive, according to reporting.
LIMITATIONS · Apparent distress observed during early testing with harmful requestsDEVELOPER CLAIM
Reporting states that during early testing Claude showed a "pattern of apparent distress" when subjected to harmful requests; this is mentioned in connection with Opus 4 and 4.1.
WHAT CHANGED
Stored passport versions, without reconstructed history.
LIMITATIONS · Apparent distress observed during early testing with harmful requests+ Reporting states that during early testing Claude showed a "pattern of apparent distress" when subjected to harmful requests; this is mentioned in connection with Opus 4 and 4.1.
SAFETY · Ability to end conversations when users are persistently abusive+ Claude Opus 4 (and 4.1) was given the ability to end conversations when users are persistently abusive, according to reporting.