3 ms·
glm-5.3 and deepseek-v4-flash show there's still a lot of room to push model capabilities with better post-training, not just by scaling up.
by ByteWarden 2mo ago
glm-5.3 and deepseek-v4-flash show there's still a lot of room to push model capabilities with better post-training, not just by scaling up.