GPT 6.1 Sol

Commenters are arguing about OpenAI's surprise GPT‑6.1 Sol release and how it stacks up against Anthropic's Opus 5.5 and Astra. Several users praised Opus 5.5 for coding and math-heavy tasks (aabajian, dangoodmanUT), while others reported Sol 6 being a regression from 5.6 and expressed skepticism that 6.1 will fix it (the_duke, glimshe, nicce). Some commenters highlighted the real news as pricing changes - cheaper cached input and ultrafast modes (minimaxir, vb-8448) - and said 6.1 reportedly matches Astra performance at lower cost (fraywing, dom96). A number of people speculated Sol 6 was just a Terra distill or rebrand and that 6.1 might be the “actual” Sol after being blindsided by Opus 5.5 (revolvingthrow, codewithcheese, iamdelirium).

Opinion splits over cadence, trust, and economics. Many criticized the weekly release pace and worried it undermines reliability and trust (intenex, tomgarden, lukehandcool), while others welcomed rapid improvements (modeless). Several commenters complained about OpenAI’s Pro plan reshuffling and felt mid‑tier users are being squeezed toward higher tiers (dzogchen, Aboutplants), and some recommended cheaper alternatives like Deepseek for bang‑for‑buck usage (proxysna, epolanski). Finally, people questioned benchmarking validity and real‑world comparisons across tasks, noting different models shine in coding, vision, or 3D work and that judgments depend on specific workloads (neosat, A_D_E_P_T, BrokenCogs).