4 ms·
This model has: 1. On par or better performance than Llama-3-70B-Instruct 2. A much more comfortable context length of 128k (vs the tiny 8k that really hinders
by aubanel 2y ago
This model has:
1. On par or better performance than Llama-3-70B-Instruct
2. A much more comfortable context length of 128k (vs the tiny 8k that really hinders Llama-3)
These 2 feats together will probably make it the first serious OS rival to GPT-4!