Cursor users notice possible improvement in 4.8
A experience says the 4.8 model felt stronger after being brought back into a setup as a and chat agent. It was used through the API for product-management-related . Compared with one week earlier, it seemed more stable and more capable.
It more consistently remembered to document decisions within the same conversation. It also pushed back and reasoned through issues instead of just agreeing. Its seemed to hold up longer as the grew.
The open question is whether Anthropic quietly improved Opus after API customers lost access to a top-tier model option.
Key points
- The 4.8 model felt better than it did about one week earlier.
- It remembered to document decisions in the same conversation more often.
- It seemed less like an echo chamber and more willing to reason through issues.
- It appeared to stay useful for longer inside a growing .
- There is speculation that Anthropic may have quietly improved Opus.