@LexnLin Duh, it's a much smaller model, no one realistically expects it to be that strong, hell, the scores are great considering Kimi is 1T in size, Mini Max is still very much impressive, looking forward to testing the weights when they drop
@BoilingOverIt@theo No, of course not, but you might be one if you expect it to last forever, you surely know that it's heavily subsidised by god knows what deal, and it's bound to fall sooner or later
@melvynx No one really takes benchmarks seriously or at face value, minimax is still 20 times cheaper, and performs well for it's size, and is open source, a great boon for the community, keep paying for GPT no one is stopping you
@drdomicile@Angaisb_ It's not, you claimed this is their "first release" when in reality the MAI team has been hard at work for years, I bet you didn't even know that lol
@drdomicile@Angaisb_ No it's not? They've been releasing fine tunes of DeepSeek for a while now, remember MAI DS? They clearly dabbled a lot and tried to create a great training pipeline, the model still sucks unfortunately
@spinelessaisha@AeroLOZ97@reikodev@grucciflipfl0ps The same social media that you're in right now? And that has helped you spread your opinion? Everything has it's downsides, but also the positive and impactful aspects, same for LLMs
@Grummz It's mid, but still a nice project, I feel every dev has (vibe)coded their own agentic frontend at some point, it's just that it seems pewd is famous thus it's getting lots of traction, I'm a 3D artist yet I still also coded my private frontend lol