Key facts
- DeepSeek has released the general availability version of its V4 Pro model, now designated as the 0813 build.
- Internal benchmarks indicate the new DeepSeek V4 Pro model is, on average, 5.3% behind Anthropic's Claude Fable 5 across nine agent benchmarks.
- The cost to use DeepSeek V4 Pro is approximately $0.435 per million input tokens and $0.87 per million output tokens.
- Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens.
- The price difference represents a 4,600% cost advantage for DeepSeek V4 Pro.
DeepSeek has quietly updated its flagship large language model, V4 Pro, to a new build (0813) that is now generally available. This new version, which had been in preview since April, shows marginal performance gains compared to its predecessor but trails Anthropic's Claude Fable 5 by an average of 5.3% across nine agent benchmarks, according to DeepSeek's own published data. However, the cost disparity is immense: DeepSeek V4 Pro is priced at approximately $0.435 per million input tokens and $0.87 per million output tokens, while Claude Fable 5 commands $10 per million input and $50 per million output. This makes DeepSeek's offering roughly 4,600% cheaper. The company released the finished version without an announcement, with the update only evident on its API pricing page. Independent verification of the 0813 build is pending, as previous benchmarks were conducted on an unfinished version. The trend of Chinese open-weight models offering performance close to frontier models at a fraction of the cost continues, with DeepSeek's weights available under an MIT license on Hugging Face.