DeepSeek V4 is not in the same league as GPT-5.5 or Opus 4.7. Benchmarks put it slightly below both of those, roughly on par with Opus 4.6. You can check the numbers yourself here: https://www.reddit.com/r/singularity/s/jIsNEK6Rrm
And yes, benchmarks only tell part of the story. In real-world usage, my experience is that V4 performs at around GPT-5.2 level, solid, consistent, and the best open-source model available right now, but doesn't quite reach Opus 4.6 in practice either.
But here's why none of that really matters, whether DeepSeek beats Claude or GPT, it achieves this level of performance with only 20% of the hardware requirements, while being fully open-source and free to download. For now running it locally is extremely demanding tho, out of reach for most people. I've been accessing it through atlascloud, and the experience has been great. At its price point, nothing else comes close. This is genuinely the cheapest SOTA model on the market by a significant margin.
[link] [comments]
