DeepSeek is really far behind on WeirdML (non-agentic, but agentic results are correlated with the normal mode). It it cheap though. But not as cheap as GPT-5.4-nano. I would be surprised if they didn't match Kimi in the update
Gemma is actually insane it also shows that this is an outlier capability. Generally, Gemma is nowhere near V4-Pro.
DeepSeek is really far behind on WeirdML (non-agentic, but agentic results are correlated with the normal mode). It it cheap though. But not as cheap as GPT-5.4-nano. I would be surprised if they didn't match Kimi in the update
DeepSeek is really far behind on WeirdML (non-agentic, but agentic results are correlated with the normal mode). It it cheap though. But not as cheap as GPT-5.4-nano. I would be surprised if they didn't match Kimi in the update
Users feel optimistic about ongoing AI competition on the WeirdML leaderboard because they expect stronger future performances from trailing models like DeepSeek.