MiniMax teases M3 model with 15.6x faster decoding speed boost
MiniMax has announced its upcoming M3 model, which boasts a 15.6x faster decoding speed compared to its predecessor, M2. The new model utilizes a technique called MiniMax Sparse Attention to improve efficiency while maintaining output quality. However, details regarding the model's parameters, licensing, and release timeline remain undisclosed.
- ▪MiniMax's M3 model achieves a 15.6x faster decoding speed and 9.7x faster prefill speed compared to M2.
- ▪The M3 model employs MiniMax Sparse Attention, which selectively focuses on relevant data blocks for improved efficiency.
- ▪MiniMax was founded in early 2022 and went public on the Hong Kong Stock Exchange in January 2026.
Opening excerpt (first ~120 words) tap to expand
MiniMax teases M3 model with 15.6x faster decoding speed boost The Shanghai-based AI firm's upcoming sparse attention architecture promises dramatic efficiency gains that could ripple through decentralized inference and crypto-native AI projects. Share Add us on Google by Editorial Team May. 27, 2026 window.sevioads = window.sevioads || []; var sevioads_preferences = []; sevioads_preferences[0] = {}; sevioads_preferences[0].zone = "01f21ccf-2092-46b1-9ac7-8c44cc782e0f"; sevioads_preferences[0].adType = "native"; sevioads_preferences[0].inventoryId = "c5700508-581b-472c-8fdd-a931cdbfc8e1"; sevioads_preferences[0].accountId = "1e47efc1-ec2d-4fca-a8b9-354e249e5095"; sevioads.push(sevioads_preferences); MiniMax, the Shanghai-based AI lab backed by Tencent, Alibaba, and miHoYo, just dropped a…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Crypto Briefing.
Discussion
0 commentsMore from Crypto Briefing
