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
Launch
CandyDrop
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
Promotions
AI
Gate AI
Your all-in-one conversational AI partner
Gate AI Bot
Use Gate AI directly in your social App
GateClaw
Gate Blue Lobster, ready to go
Gate for AI Agent
AI infrastructure, Gate MCP, Skills, and CLI
Gate Skills Hub
10K+ Skills
From office tasks to trading, the all-in-one skill hub makes AI even more useful.
GateRouter
Smartly choose from 40+ AI models, with 0% extra fees
DeepSeek Web Version Quietly Launches "Fast" and "Expert" Dual Modes, Introducing Product Layering for the First Time
Community analysis shows that behind Quick Mode is DeepSeek 3.2, with a knowledge cutoff of July 2024. Expert Mode points to a newer model, possibly an early version of V4. In user tests, Expert Mode performs significantly better on deep reasoning tasks such as physical simulation and mathematical reasoning, but the gap versus Quick Mode is limited on simpler tasks like creative writing. Some testers believe that Expert Mode’s current routing is still to some version of V4 Lite, and that the full version of V4 will require more waiting. In the front-end code, a third option that has not yet been launched—“Vision Mode”—was also found. However, reverse engineering indicates that it is not an independent model; instead, a visual understanding parameter is enabled within Quick Mode.
This is DeepSeek’s first time doing product tiering since it went viral at the beginning of last year. Previously, the web interface was completely free and had no functional distinctions; it would route users based on their needs to different model entry points, which in itself is a compute-scheduling strategy. And once this architecture is in place, technically there is no longer any barrier to introducing a paid system later or imposing mode-based limits.