xynelius

I think it's safe to assume every single model is benchmaxxed. I mean why wouldn't you include training data from common benchmarks? Not only is a higher score good marketing, but these benchmarks are supposed to represent common tasks so training for those should in principle improve performance across a wider range of tasks.

I was curious how much energy is actually needed to power these datacenters, so I did a little bit of math.

Looking at Nvidia revenues in the past few years, there's maybe $300 billion worth of GPUs currently deployed in the U.S. The B200 costs ~$40k, so we have 7.5 million B200-equivalents, which draw 1000W. Running these at full capacity requires 66 TWh a year, or ~1.5% of total current U.S. electricity consumption. Maybe a bit more to account for inefficiencies, cooling, and other components, but not more than ~2.5% total I would guess.

So it's not that much in reality, but will definitely grow fast.

The Doom demo looks impressive but was it a fine-tuned model? It's the difference between a cool demo and revolutionary tech.

If that wasn't impressive enough, it's actually ~60x cheaper if you take into account the typical cache-read/input/output split in agentic coding, and the deep discount for cache reads offered by DeepSeek. Opencode has some public data on the typical split [1]:

For DeepSeek V4 Pro the typical split is 750 in, 290 out, 82k cached.

Cost per request for V4 Pro: $0.000875 per request.

Equivalent Opus cost (w/o taking into account cache write costs): $0.052 per request.

[1] https://opencode.ai/docs/go/#usage-limits

It's not bad that your desires are influenced by what others want (it's inevitable, we're social animals), but understanding how those influences work is part of understanding your desires and what's behind them.

Personally, I would be miserable if the people around me didn't approve of me, even if it's for their own selfish reasons.