Paid for Sonnet 4.6 but Response Quality Has Dropped Significantly — Starting to Suspect It's Actual

[kui niu]

I'm a paying RovoDev customer with valid billing on file, and I'd like to raise a serious concern.

I've been using the Claude Sonnet 4.6 tier, but the responses I'm getting feel noticeably weaker in reasoning, depth, and overall quality — more consistent with Claude Haiku 4.5 than Sonnet 4.6. I strongly suspect I'm not being served the model I'm paying for.

This is a significant issue. I'm paying a premium for Sonnet 4.6 and expect to receive exactly that. If the correct model is not being delivered as advertised, I'd like to request a full refund.

[Dr Valeri Colon _Connect Centric_]

Hi @kui niu you’re not alone—this has come up before, but it’s rarely a confirmed “model swap.” In Rovo Dev:

• Model routing can vary by context, limits, or fallback behavior
• Quality changes can come from prompt structure, tool use, or updates (like your version bump)

There’s no user-facing way to verify the exact model per response today. If this started after 0.13.57, it could be a regression—worth raising with support, including example prompts + outputs so they can trace it.

[kui niu]

I noticed this issue started after I upgraded to version 0.13.57. Has anyone else experienced the same problem after updating to this version?