Cerebras Code now supports GLM 4.6 at 1000 tokens/sec