It doesn't. It's called preserved reasoning and every recent reasoning model does it
Sorry, I've realised I was only partially correct.
Gemini[0] for example passes along a snapshot of the reasoning state but it's not the equivalent to keeping all the reasoning tokens in the context.
[0] https://ai.google.dev/gemini-api/docs/thinking#signatures
Edit: Apparently it does take the same space in the LLM latent space so I was wrong.
Sorry, I've realised I was only partially correct.
Gemini[0] for example passes along a snapshot of the reasoning state but it's not the equivalent to keeping all the reasoning tokens in the context.
[0] https://ai.google.dev/gemini-api/docs/thinking#signatures
Edit: Apparently it does take the same space in the LLM latent space so I was wrong.