Jarvislabs 上线 V4.1 Flash：缓存输入 $0.003/1M

Vishnu - Jarvislabs.ai @vishnuvig原推发布于 2026.10.25

在 X 上打开原推

Most of what we pay for in a coding agent session is input tokens. The agent re-reads the same repo context every turn, so cached input pricing matters a lot more than output pricing. DeepSeek V4.1 Flash (1017) is now live on Jarvislabs Token Factory, with a 1M context window. Input: $0.07 / 1M tokens Cached input: $0.003 / 1M tokens Output: $0.49 / 1M tokens Speed: 180 to 200 tok/s Reading a 100k-token codebase 100 times costs about 3 cents in cached input. It works with the OpenAI SDK, Claude Code, Codex and OpenCode