5 ms·
If you believe benchmarks then competitors caught up with GPT-4 months ago, but this is certainly not true for my usecases. In particular, I find GPT-4 halluci
by sebzim4500 3y ago
If you believe benchmarks then competitors caught up with GPT-4 months ago, but this is certainly not true for my usecases.
In particular, I find GPT-4 hallucines less than Gemini and much less than Claude 3.
- hackerlight 3y agoAre you using Claude 3 with low temperature and low top p? That's recommended for code generation.