Huh? They're several times cheaper than SOTA models at market rate prices.

pants2 · 2026-05-28T19:59:09 1779998349

If you are only looking at US hosting providers, models from US labs easily meet or beat models from Chinese labs on the same intelligence level. I'm not comparing DeepSeek with Opus because those are on different levels of performance.

slopinthebag · 2026-05-28T20:31:48 1780000308

Deepseek v4 Pro on US hosting is like 1.5x cheaper and 5x cheaper on input/output compared to Sonnet, and that's not even a fair comparison because Deepseek is much stronger than Sonnet. It's more reasonable to compare with Opus 4.5, which is much more expensive.

pants2 · 2026-05-28T23:29:08 1780010948

Sure but you can also look at Grok 4.3, which is smarter and faster than DeepSeek at the same price point.

slopinthebag · 2026-05-29T16:01:15 1780070475

I doubt that is the case

pants2 · 2026-05-29T20:42:11 1780087331

Grok 4.3 is smarter, cheaper, and faster than DS4 according to Artificial Analysis and OpenRouter:

DeepSeek v4 Pro

Intelligence: 52

Cost to run AA benchmark on Fireworks: 187M tokens @ $0.79/M blended cost = $147.73

Speed: 43 tok/s

Grok 4.3

Intelligence: 53

Cost: 88M tokens @ 0.64 blended = $56.32

Speed: 108 tok/s