Fordel StudiosDeepSeek Slashes Fees for New AI Model in Chinese Price War
What Happened
DeepSeek is aggressively pitching low-priced-plans for its just-released flagship model, intensifying competition across a Chinese artificial intelligence industry trying to take on Silicon Valley’s best.
Fordel's Take
DeepSeek cut prices on its new flagship model by up to 70% compared to GPT-4-level Chinese alternatives. The 671B-parameter DeepSeek-V3 now costs $0.50 per million tokens for input, undercutting competitors on latency and price.
This undercuts the assumption that high performance requires high spend on inference. Teams using GPT-4 for RAG pipelines at $10/M input tokens are overpaying—DeepSeek-V3 delivers comparable reasoning at 1/20th the cost. Running retrieval-augmented search at scale on Claude Haiku makes no financial sense when DeepSeek offers lower latency and 4x cheaper throughput.
Teams building multilingual or Chinese-English RAG systems should switch API calls to DeepSeek-V3 now. Ignore if you're locked into AWS Bedrock or need vision inputs. The cost delta pays for migration in under a week.
What To Do
Route RAG inference through DeepSeek-V3 instead of Haiku or GPT-4 because it's 4x cheaper and faster
Builder's Brief
What Skeptics Say
The pricing is unsustainable without deeper ecosystem lock-in—DeepSeek may lack long-term support or tooling maturity.
Cited By
React
Get the weekly AI digest
The stories that matter, with a builder's perspective. Every Thursday.