You can buy a pair of DGX Sparks and run Deepseek V4 Flash at ~60-70TPS (once DSpark support matures over the next few days).
replyThat will get you a near-frontier experience. DSv4 Flash launched in April with capabilities on par with GLM 5.0, which launched in February.
I really think giving it a year for the hardware market to come back to earth and spending a fraction of that for API access to the same models is a better use of the money.
replyImplicit in your answer is the belief that they will come back to earth. I wonder how realistic that belief is.
replyWe have decades upon decades of hardware getting dramatically cheaper year over year for the same performance, and ~1 year of the inverse due to dramatic buildout for AI.
replyIt's a surprising example of the recency bias to me to assume anything other than the market returning to its historic norm, even if the AI buildout doesn't slow, producers will scale factories to meet that demand.
I look forward to re-evaluating this statement in, what do you say, 12 months from now?
reply