A user successfully ran Claude Code, a tool designed for Anthropic's models, on DeepSeek's V4.1 Flash model. Despite DeepSeek's off-peak pricing of approximately $0.02 per task, Claude Code incorrectly estimated the cost at over $1.00 per task by defaulting to Claude Opus 5 pricing for the unrecognized DeepSeek model. The user confirmed that the coding task was completed accurately across three separate runs, with the model even logging changes to a file despite instructions not to alter anything beyond the requested task. AI
IMPACT Highlights potential cost discrepancies and integration issues when using third-party models with specialized AI tools.
RANK_REASON User testing of an existing tool (Claude Code) on a different model provider (DeepSeek).
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →