DeepSeek's price increase lands this weekend. What are you actually switching to?

DeepSeek raises prices at 16:00 UTC on Sunday the 16th. Peak-hour output on V4-Flash goes from $0.28 to $1.32 per million tokens, so roughly a 4.7x jump, and across the V4 line the reported increases run from about 50% to over 1,100% depending on the model, whether it's input or output, and what time of day you're calling it. That still leaves it cheaper than most of the frontier APIs, so this isn't a "DeepSeek is over" post. But a lot of people picked it specifically because the price made a whole category of thing viable: batch jobs, multi-call agent loops, anything where you're burning tokens on volume rather than on difficulty. If a workflow only worked at $0.28, that's when you find out. Curious what people are actually doing about it rather than what the benchmarks say. If you're on DeepSeek in production, does the new pricing change anything for you, or was the cost never the binding constraint? Has anyone moved a real workload to one of the other cheap hosted options and measured the quality difference honestly, including the cases where it got worse? And for anyone who's gone local instead, at what monthly volume did that actually start making sense, hardware included? Concrete numbers more useful than impressions here. "It's fine" doesn't help anyone planning a migration.