DeepSeek Models 17-100x Cheaper Than Moonshot, Dominating Long-Context Costs

teortaxesTex · x · 2026-07-31

The post discusses the massive cost advantages of DeepSeek's models. The author points out that while Moonshot might maintain higher margins, DeepSeek's models are inherently vastly cheaper—especially for long contexts—costing up to 50x less per session and capable of running on commodity hardware.

In the quoted discussion, a user speculates that DeepSeek's massive performance jump might simply come from generating more training environments and performing more RL steps. The author agrees, emphasizing that DeepSeek's recipe is no worse than Moonshot's, yet their models are 17 to 100x cheaper.

Related event: DeepSeek V4-Flash Benchmarks Surpass Pro, Sparking Cost-Efficiency Debate(18 posts)→

Original post →

More from Models

Models channel →