I think the jury is still out on K3/Q3.8 and if they are equivalent to Opus or Fable. Benchmarks have gotten incredibly murky and I've tried models that are "the same as X model" and been unimpressed. I've tried other models with Claude Code and with tools like OpenCode or Pi and nothing has really come close to Claude Code using Anthropic's models (mostly Opus 4.8).

I'm not saying these other models are trash, just that I'm not quite ready to put them on equal footing to Anthropic or OpenAI models.

I think the future of LLM coding (and more) is probably routers that decide on a per-task basis which model to route to. Know when to use Fable/Opus and when to fall back to DeepSeek/Kimi/Qwen, even all the way to local models. Of course that's not in the frontier lab's interest but it feels like there is a lot of low-hanging fruit there. I hate that right now it's pretty much "just use the same model for planning/execution/etc" (without standing on your head).