Other
LLaDA-2.1-mini
Timeline (last 90 days)
Oct 17
On LLaDA-2.1-mini, ORL requires only 4.2M self-generated training tokens and 36 H100 GPU hours of post-training to obtain up to a 3x throughput improvement with comparable quality.
Unverified50%
Oct 17
A single RTX 5090 with LLaDA-2.1-mini + ORL reaches 2,788 tokens/s at NVFP4 precision (single-request decoding, batch size 1).
Unverified50%
All Facts (2)
Unverified
On LLaDA-2.1-mini, ORL requires only 4.2M self-generated training tokens and 36 H100 GPU hours of post-training to obtain up to a 3x throughput improvement with comparable quality.
50%UnverifiedA single RTX 5090 with LLaDA-2.1-mini + ORL reaches 2,788 tokens/s at NVFP4 precision (single-request decoding, batch size 1).
50%