Anthropic Tests Claude 4 Models with Internal Users

Anthropic is approaching the public release of its next-generation Claude 4 model family, which includes Claude Sonnet 4 and Claude Opus 4. These models have been spotted in early testing rounds, with access currently limited to a select group of internal testers known as 'Ants'. The models are not yet production-ready and are subject to strict rate limits, indicating they are still under active evaluation.

The Claude 4 models are part of Anthropic's broader testing initiative, Claude Neptune, which aims to enhance model integrity and prevent prompt exploits. Notably, Claude 4 is classified under the ASL-3 safety tier, indicating higher capability and potential misuse risk compared to the current ASL-2 models.

The current testing phase, described as 'friends and family' testing, suggests that a broader developer or public preview may be imminent. If Anthropic's past release patterns hold, Claude 4 could officially launch around June. For now, the models remain in the final rounds of internal vetting, with users eagerly awaiting their enhanced autonomous reasoning features.