Source
cache_creation
Happens once per session the first time Claude processes a prompt — and again after 5 min of inactivity (cache expires, so the next message re-pays the write cost). Costs $3.75/MTok — 25% more than regular input. Pays off only if enough reads follow before the next expiry.
cache_read
Reusing a cached context costs $0.30/MTok — 90% cheaper than fresh input. Every read earns back the write premium. This is the savings mechanism.
break-even
Need ~0.28 reads per write before caching saves money. Below this threshold you spend more than you save. Efficiency < 0.28× = poor status.
short sessions
Cache expires after 5 min of inactivity. Short sessions write cache that is never read back. Keep sessions long and focused to maximise read reuse.
/* Net Savings Formula (Sonnet pricing) */
savings = ( cache_read × 0.9 − cache_creation × 0.25 ) × $3.00 / 1,000,000
poor → efficiency < 0.28× (paying more than saving)
warn → efficiency 0.28—1.5× (near break-even)
good → efficiency > 1.5× (clear net positive)
savings = ( cache_read × 0.9 − cache_creation × 0.25 ) × $3.00 / 1,000,000
poor → efficiency < 0.28× (paying more than saving)
warn → efficiency 0.28—1.5× (near break-even)
good → efficiency > 1.5× (clear net positive)
—
Efficiency
—
Status Breakdown
Est. Savings
—
Cache Hit %
—
Sessions
—
Est. Cache Savings
—
vs. uncached baseline
Cache Hit %
—
of input served from cache
Cache Efficiency
—
reads per write token
Status
—
—
Insights & Recommendations
Loading recommendations…
By Project
—Project-level cache breakdown
Loading…
| Project | Sessions | Cache Read | Cache Write | Hit % | Efficiency | Est. Savings | Status |
|---|---|---|---|---|---|---|---|
| Loading… | |||||||
By User
—User-level cache breakdown
Loading…
| User | Sessions | Cache Read | Cache Write | Hit % | Efficiency | Est. Savings | Status |
|---|---|---|---|---|---|---|---|
| Loading… | |||||||
By Team
—Team-level cache breakdown
Loading…
| Team | Sessions | Cache Read | Cache Write | Hit % | Efficiency | Est. Savings | Status |
|---|---|---|---|---|---|---|---|
| Loading… | |||||||