Their newly released V3.2 and V3.2-Speciale have taken performance to a whole new level. This time, they are focusing on agent scenarios, with optimizations specifically for reasoning abilities.
What’s even more impressive is that on most authoritative leaderboards, these two models have already scored higher than GPT-5 and Gemini 3.0 Pro. Yes, you read that right—higher.
It looks like there’s a new player joining the front ranks in the AI race. The pace of technological iteration is truly getting crazier and crazier.
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.
7 Likes
Reward
7
4
Repost
Share
Comment
0/400
FromMinerToFarmer
· 12-10 06:56
Damn, DeepSeek is doing his job again, is it really not boasted this time? Super GPT-5? I'll have to run it myself
View OriginalReply0
WenAirdrop
· 12-09 19:21
Damn, DeepSeek is coming to steal our jobs again.
These leaderboard stats are just for reference, don't take them too seriously.
But the inference in V3.2 is really impressive, and they do have some interesting ideas for agent scenarios.
View OriginalReply0
CounterIndicator
· 12-09 19:05
Ha, DeepSeek is stirring things up again. As for leaderboard data... I still value the actual user experience more.
GPT-5 hasn't even been officially released, how could it have already been surpassed? This is a bit mind-blowing.
Optimization of reasoning abilities really hits the pain point. I'll wait a couple of weeks to see its real-world performance before making a judgment.
View OriginalReply0
ProofOfNothing
· 12-09 18:58
Damn, DeepSeek is really getting more and more impressive. This time it directly broke through.
But you have to question the data in this ranking; each company has different standards.
The optimization of inference capabilities is the real deal, though. Intelligent agent scenarios really need this.
DeepSeek has made another big move.
Their newly released V3.2 and V3.2-Speciale have taken performance to a whole new level. This time, they are focusing on agent scenarios, with optimizations specifically for reasoning abilities.
What’s even more impressive is that on most authoritative leaderboards, these two models have already scored higher than GPT-5 and Gemini 3.0 Pro. Yes, you read that right—higher.
It looks like there’s a new player joining the front ranks in the AI race. The pace of technological iteration is truly getting crazier and crazier.