Similar story here. DS models are absurdly good value for mid-end tasks. I've found DSv4 Flash to be ~10% the cost of GPT-5.4-mini/Claude Haiku at similar performance.
replyWe used to pay OpenAI >1m$/month for fraud classification, NER, etc. Sadly the US companies no longer care about non-coding-agent uses.
I imagine uptake will continue to increase as the corporate infra improves. Right now it's still bad - for example, AWS Bedrock is awful, models are months late and implemented with basic errors. Google Vertex is even worse. Finding a decent provider is the hardest part.
Yeah we don't use it for coding either. Basically text analysis within specific constraints. It works beyond my expectations.
replyWhere is your DeepSeek model hosted?
replyDeepseek directly but we can switch to openrouter and a USA host at any time. Again it's just a configuration value - no code changes.
reply