The Chinese AI lab DeepSeek has launched an upgraded version of its V4 Pro model with a 1 million token context window. The company claims that Anthropic’s Claude Fable 5 only outperforms it by 5% on key benchmarks while charging approximately 45 times more for API access. DeepSeek offers pricing of $0.43 per million input tokens and $0.87 per million output tokens, compared to $10 and $50 respectively for Claude Fable 5. Claude Fable 5, however, dominates general intelligence tests with a score of approximately 59.9 versus DeepSeek’s 44.3. DeepSeek’s Chinese origin remains a potential barrier to enterprise adoption in regulated industries.
Source: Read the original article
