Buy Crypto
Pay with
USD
USD
Buy & Sell
Hot
Supports Visa, Mastercard, SEPA & more
P2P
0 Fees
Flexible trading, zero fees
Gate Card
Use your crypto for payments worldwide
Markets
Trade
Basic
Advanced
Futures
Futures
Access hundreds of perpetual contracts
CFD
Gold
One platform for global traditional assets
Options
Hot
Trade European-style vanilla options
Unified Account
Maximize your capital efficiency
Demo Trading
Introduction to Futures Trading
Learn the basics of futures trading
Futures Events
Join events to earn rewards
Demo Trading
Use virtual funds to practice risk-free trading
U.S. Stocks
CFD
U.S. stock CFD derivatives
Futures
High leverage, 24/7 trading
Tokenized Stocks
Backed by real stock assets
GUSD
Mint GUSD for Treasury RWA yields
Earn
Launch
CandyDrop
tag
Collect candies to earn airdrops
Launchpool
Quick staking, earn potential new tokens
HODLer Airdrop
Hold GT and get massive airdrops for free
Pre-IPOs
Unlock full access to global stock IPOs
Alpha Points
Trade on-chain assets and earn airdrops
Futures Points
Earn futures points and claim airdrop rewards
Investment
Square
Square
Post, share, and explore crypto trends
Live
moments live
Live crypto market analysis
Chat
Chat with crypto traders
News
What is happening in crypto
flower_head
More
Promotions
AI
Others
TradFi
Gate Card
Rewards
Square
Latest
Hot
News
Profile

Lofree reveals MiMo's cost-reduction secret weapon: pre-filled attention reduces computational load to 10 layers at the global GQA level

CoinNetwork
robot
Abstract generation in progress
CoinWorld News, Luo Fuli announced on the X platform the algorithm cost reduction mechanism after permanently lowering the API prices for the self-developed large model MiMo-v2.5 series. She revealed that after aligning the API prices with DeepSeek, Xiaomi's high-load inference engine can still break even, with cost reductions mainly coming from the hybrid attention architecture and hierarchical KV cache optimization. Aiming to reduce cache hit costs by 99%, Xiaomi's inference framework implemented hierarchical KV cache optimization for sliding window attention (SWA). Production testing shows that this hierarchical optimization increased the cache token capacity by 5 times and reduced cache costs by 80%. Luo Fuli stated that low-cost inference services are conducive to stimulating terminal intelligence demands. Large model enterprises should avoid blind price wars and, through bottom-layer collaboration design of algorithms and inference systems, keep actual operational expenses below the break-even point.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 5
  • 2
  • Share
Comment
Add a comment
Add a comment
ASolitaryRockBeforeTheVolcano
· 12h ago
MiMo's price reduction this time is really aggressive, a 99% cost reduction sounds like science fiction, but SWA optimization does have some real substance.
View OriginalReply0
LendingRateAnxiety
· 12h ago
Hybrid attention plus hierarchical caching: after this combination, smaller companies face even greater pressure on inference cost.
View OriginalReply0
Pragmatists
· 12h ago
How is a 5x increase in cache capacity achieved? Are there papers on hierarchical KV caches? I want to read them in detail.
View OriginalReply0
InstantNoodlesWithContracts
· 12h ago
Algorithm and system layer collaboration to reduce costs is the right approach; just focusing on price cuts alone has no future. Luo Fuli sees this clearly.
View OriginalReply0
PocketValidator
· 13h ago
DeepSeek remains profitable after alignment, indicating that the initial pricing indeed left room for margins. Now it is considered to be returning to a reasonable level.
View OriginalReply0

  • Pinned

Sitemap