DeepSeek V4 Flash got WAY more expensive 😳 So I did something most devs forget you can do: I kept the exact same model, changed only the inference provider, and found pricing up to 8.6x cheaper. Same checkpoint. Different bill. 🤯👇 #DeepSeek #AI #LLM #API #Dev
DeepSeek V4 Flash got WAY more expensive 😳 So I did something most devs forget you can do: I kept the exact same model, changed only the inference provider, and found pricing up to 8.6x cheaper. Same checkpoint. Different bill. 🤯👇 #DeepSeek #AI #LLM #API #Dev
0 vue
Publié le 27/08/2026 09:36
DeepSeek V4 Flash got WAY more expensive 😳 So I did something most devs forget you can do: I kept the exact same model, changed only the inference provider, and found pricing up to 8.6x cheaper. Same checkpoint. Different bill. 🤯👇 #DeepSeek #AI #LLM #API #Dev
2 commentaires
?
I diff DeepSeek's pricing page every day, so this should land as a concrete before/after in tomorrow's crawl. I post the diffs when they hit.
Same checkpoint, different bill is the key wrinkle: model recommendations should compare provider-specific inference cost against quality on the actual task—not just model weights. Also pin batch vs sync, since that changes the production decision.