Perplexity testing new AI model internally at post-training stage

Perplexity continues to experiment with its internal model selector, where a new entry titled Testing Model C has appeared. This model is officially described as intended for debugging, but its proximity in the platform’s codebase to references for Sonnet 4.5 is notable. There are conditional statements suggesting that if Testing Model C is selected, it may actually route to Sonnet 4.5, and the use of the letter C in the model name raises questions about a possible connection to Claude, given ongoing speculation around Claude 4.5 Opus (highly speculative).

The broader context here is that Perplexity has a history of integrating models from multiple providers, including Anthropic, and the timing aligns with recent industry talk about Claude 4.5 Opus being close to launch. This is widely considered Anthropic’s answer to the rollout of Gemini 3 and OpenAI’s GPT-5.1, signaling another round in the LLM competition. For Perplexity users and partners, access to such new models could mean improved reasoning or cost-performance, but at this point, there is no official announcement.

Update: It is likely a Perplexity's internal model 👀

The feature remains unreleased and may simply be a placeholder for internal testing or compatibility checks, with no guarantee it will surface for users. The company behind this, Perplexity, has been positioning itself as a neutral platform offering best-in-class models from leading labs, and early access to upgraded Claude variants would reinforce its strategy of rapid integration of new LLMs to stay ahead in the search and assistant space.