[Feat] Support time-based (peak/off-peak) pricing for DeepSeek V4 Pro and Flash

Problem

DeepSeek switched to peak/off-peak billing on 2026-08-16. Prices now depend on the time of day, but models.dev only stores a single flat [cost] block per model, so the current entries are stale and there is no way to represent the real pricing.

Peak hours: 01:00–04:00 UTC and 06:00–10:00 UTC (7 h/day). All other hours are off-peak (half price).

New pricing (USD per 1M tokens)

DeepSeek V4 Pro

DeepSeek V4 Flash

(Reasoning tokens are billed at the output rate; cache read/write rates also vary by peak/off-peak.)

Current (stale) repo state

• deepseek-v4-pro.toml: input 0.435 / output 0.87 / reasoning 0.87 / cache_read 0.003625
• deepseek-v4-flash.toml: input 0.14 / output 0.28 / reasoning 0.28 / cache_read 0.0028

References

• DeepSeek API docs - Models & Pricing: https://api-docs.deepseek.com/quick_start/pricing

Start by locating how the flat [cost] block is represented and how deepseek-v4-pro.toml and deepseek-v4-flash.toml are loaded. Check the DeepSeek pricing reference for peak and off-peak rates, then update the model data so both periods and cache rates are representable and the two stale entries use the new prices.