Comparing Prime Mega MoE with our prime-rl baseline:
- Training steps on Qwen3-30B-A3B are 34% faster
- Training steps on Qwen3-235B-A22B are 32% faster
Prime Intellect's Mega MoE trains Qwen3 models 32–34% faster
AIPrime Intellect says its Prime Mega MoE cuts training step time by 34% on Qwen3-30B-A3B compared with its prime-rl baseline. On Qwen3-235B-A22B, the company reports a 32% faster training step.
@PrimeIntellect
Part of a thread · earlier post
Source details
Prime Intellect · x.comPublished · added here
Original title: Comparing Prime Mega MoE with our prime-rl baseline: