The “stolen CoT” paper has some thinking traces of closed frontier models.
The ‘summarisers’ mask a lot of raw thinking you’d probably be frustrated by too.