GPT-6 Sol's price cut means it should be compared to Sonnet, not Opus
TraditionalHome8852 · reddit · 2026-09-23
The poster argues from pricing logic: GPT-6 Sol dropped to $2 input / $10 output, roughly matching Terra 5.6 and Sonnet 5, with no GPT-6 Terra announced — so Sol should be benchmarked against the upcoming Sonnet 5.5, not Opus 5.5.
They see OpenAI's lineup becoming clear — Luna for light high-volume work, Sol for most work, Astra for the hardest — with prices aligned to capability. On Anthropic's side, the cheaper Opus now matches or beats Fable on most tasks, making the roles hard to distinguish; Sonnet is "a lost middle child," and they predict Anthropic will eventually kill either Haiku or Sonnet.
More from Models
- LIBERO experiments show stronger reasoning in GPT-6 variants means better robot manipulation — YuXiang_IRVL · 2026-09-23
- Diffusion Gemma is undervalued: real System-1 with both intelligence and speed — bingxu_ · 2026-09-23
- Claude Opus 5.5 models individual hair strands in Blender demo — majidmanzarpour · 2026-09-23
- Opus 5.5 on Perplexity: generous credits plus automatic prompt-injection checks for skills — ChrisUniverse · 2026-09-23
- Miles Brundage jokes new models will improve on "YouTube Poop bench" — Miles_Brundage · 2026-09-23
- Robot arm tests rank GPT-6 variants: better reasoning means better manipulation at 30x the cost — YuXiang_IRVL · 2026-09-23