This is based on informally comparing samples and counting annoyingly big enough to distract the reader mistakes rather than the error rate overall.
GPT 5 seems to be worse at webnovel translation than 4o or 4.1 was, oddly enough.
This is based on informally comparing samples and counting annoyingly big enough to distract the reader mistakes rather than the error rate overall.
GPT 5 seems to be worse at webnovel translation than 4o or 4.1 was, oddly enough.