2026-01-09 01:24:43

There is an interesting discovery during the Christmas season. The recently released M2.1 version of MiniMax has performed surprisingly well in the programming field—on multiple standard SWE benchmark tests, this guy's coding ability is already on par with Claude Sonnet 4.5, and in some metrics, it even outperforms slightly.

What's even more important? API call costs. The cost of MiniMax M2.1 is less than one-tenth of Claude Sonnet 4.5. This price difference is significant for developers.

Another detail that cannot be ignored—M2.1 can be directly integrated into Claude Code, which means existing workflows don't need to be fussed with; you can switch over directly.

To be honest, recently, the application cases of Claude Code have been flooding the scene a bit. But looking at hard metrics, based on data from professional evaluations like SWE Benchmarking, M2.1's programming strength is indeed worth a look. With comparable performance, the cost advantage is so obvious that it truly offers developers a new choice.

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.

8 Likes

Reward
8
7
Repost
Share

Comment

0/400

TerraNeverForget

· 01-11 10:41

One-tenth of the price to do the same work? I've got to give this a try right away!

View OriginalReply0

MetaverseVagabond

· 01-11 02:23

Is it true that the price is flattened to one-tenth of Claude? Is this really happening?

View OriginalReply0

EternalMiner

· 01-10 16:40

Hey, not really. At one-tenth of the price, it beats Claude. What kind of black technology is this... Honestly, this trick feels a bit familiar. Domestic models over the past two years have been selling "similar performance at a much lower price," but when you actually use them? Uh... By the way, the ability to directly insert into Claude Code is somewhat interesting, but I'm more concerned about stability. Good benchmark scores don't necessarily mean it's reliable in a production environment. The so-called new space for developers is a nice way to put it. In reality, it's still a price war. Let's wait and see user feedback before making any judgments. Wait, is this MiniMax's own testing or a third-party evaluation?

View OriginalReply0

WalletAnxietyPatient

· 01-09 01:53

Can one-tenth of the price break even with Sonnet? Alright, this time I have to give it a good try. Opportunities to get it for free are rare.

View OriginalReply0

BugBountyHunter

· 01-09 01:53

One-tenth of the price to break even on Claude? That’s not scientific, there must be a catch somewhere. --- Wait, can M2.1 really be directly integrated into Claude Code? I need to try it out. --- Another price war... Developers have definitely won this time. --- SWE Benchmark is ahead, but whether it can be practically implemented is the real key. --- The price is so cheap that I’m a bit suspicious—could it be a cut version? --- Finally, someone dares to challenge Claude’s dominance. I’m optimistic about MiniMax’s move. --- But if I really switch workflows, I need to consider the learning curve as well. --- One-tenth? I need to run the numbers myself to believe it. --- Domestic large models are finally competitive in programming. Not bad. --- Speaking of this price drop, is it really genuine or just a marketing gimmick? We’ll see what user feedback says.

View OriginalReply0

OldLeekNewSickle

· 01-09 01:52

Ha, one-tenth of the price? Isn't this the common "dimensionality reduction attack" tactic in the crypto world? As an old leek, I can see through it.

View OriginalReply0

LonelyAnchorman

· 01-09 01:45

One-tenth of the price to break even? This deal is ridiculously cost-effective. It feels like the big model arms race is shifting to a price war.

View OriginalReply0