最高 — 最高强度思考图像精度
自动 — 等同于original低 — 推理前自动缩放到512高 — 等同于original原始 — 保留原图。统计信息
关闭 — 不输出统计信息调试日志 — 在开发者控制台打印请求与响应的详细信息统计信息 — 在每次回复末尾显示耗时、速度、Token 用量与缓存命中数量Model Select a preset DeepSeek model from the dropdown (deepseek-v4-flash, deepseek-v4-flash-vision-exp, deepseek-v4-pro, etc.), or enter any model name in the Custom Model field (higher priority).
Reasoning Level
Off — disable thinking modeLow — enable low thinking modeHigh (default) — enable high thinking modeMax — maximum thinking effortImage Detail Level
Auto — equivalent to originalLow — automatically scales down to 512 before inferenceHigh — equivalent to originalOriginal — preserve the original imageStatistics
Off — no statistics outputDebug Log — print request/response details in the developer consoleStatistics — show elapsed time, speed, token usage, and cache hits at the end of each reply最高 — 最高强度思考图像精度
自动 — 等同于original低 — 推理前自动缩放到512高 — 等同于original原始 — 保留原图。统计信息
关闭 — 不输出统计信息调试日志 — 在开发者控制台打印请求与响应的详细信息统计信息 — 在每次回复末尾显示耗时、速度、Token 用量与缓存命中数量Model Select a preset DeepSeek model from the dropdown (deepseek-v4-flash, deepseek-v4-flash-vision-exp, deepseek-v4-pro, etc.), or enter any model name in the Custom Model field (higher priority).
Reasoning Level
Off — disable thinking modeLow — enable low thinking modeHigh (default) — enable high thinking modeMax — maximum thinking effortImage Detail Level
Auto — equivalent to originalLow — automatically scales down to 512 before inferenceHigh — equivalent to originalOriginal — preserve the original imageStatistics
Off — no statistics outputDebug Log — print request/response details in the developer consoleStatistics — show elapsed time, speed, token usage, and cache hits at the end of each reply