Inbox

会话与 AI 草稿审阅
Sonnet for your doc pipeline — 50% of list
question置信 0.93
Gabriel →Mon 08:02

Hi Daniel — Helio's doc pipeline looks like it's running Claude Sonnet at volume. We serve the same model at 50% of list price, OpenAI-compatible, no code change. Want the math for your stack?

Daniel Park →Tue 07:55

Hi Gabriel — saw your note. We're running Claude Sonnet for our doc pipeline, ~40M tokens/mo. What's your p50 latency and do you support prompt caching? Also, is this an official API key pool?

AI 草稿 · 待审⌘↵ 发送 · e 编辑

Hi Daniel — p50 ~1.1s to first token on Sonnet, p95 2.4s. Prompt caching is supported (5m and 1h TTL). Yes — official key pool, not a reverse proxy. We can share key provenance and sign a DPA. At 40M tokens/mo you'd be at roughly $1.9k vs $3.8k at list. Want the one-pager with the math?

证据
  • 直接调 anthropic SDK
    github.com/heliolabs5 天前
  • 按 token 计费
    heliolabs.ai/pricing1 天前
  • Seed $4.5M
    sifted.eu/...52 天前
算账
Claude Sonnet · 40M tok/mo
$3.8k / 月$1.9k / 月
省 50%
知识库引用
  • KB-02 延迟 SLA 说明
  • KB-07 官 key 池来源说明
  • KB-09 DPA 模板
联系人
Daniel Park
CTO · Helio Labs
heliolabs.ai · 邮箱已验证