Start free →
Analysis window
Potential savings · next 30d
$665/ mo
across 3 recommendations · 3 high-confidence
$665 / mo high-conf
$54.70 / mo achieved ✓
Spend · 7d
$108
analyzed models
Open
3
model swaps
Achieved
$54.70
1 swap adopted
Typemodel swap · 3savings desc
Open · 3 · $665 / mo
SWAPopenai / gpt-4o → openai / gpt-4o-mini
openai / gpt-4o→openai / gpt-4o-mini
These calls average 480 prompt tokens with simple classification outputs (≤ 180 tokens). GPT-4o-mini handles this task class at ≥ 95% quality based on our benchmark suite.
SAVE / MO
$412
was $68.42 /7d
SAMPLES
1,240
~180 output tk
CONFIDENCE
high
≥$40/mo · ≥100 req
SWAPanthropic / claude-sonnet-4-5 → anthropic / claude-haiku-4-5
anthropic / claude-sonnet-4-5→anthropic / claude-haiku-4-5
Data extraction spans are structured (JSON output, fixed schema). Claude Haiku matches Sonnet on F1 score for this schema type while costing 5× less.
SAVE / MO
$185
was $29.84 /7d
SAMPLES
624
~640 output tk
CONFIDENCE
high
≥$40/mo · ≥100 req
SWAPopenai / gpt-4o → openai / gpt-4o-mini
openai / gpt-4o→openai / gpt-4o-mini
Sentiment scoring calls have consistent 1–5 integer outputs. A/B sampling on 210 requests shows identical distributions for both models on this task.
SAVE / MO
$67.30
was $9.41 /7d
SAMPLES
210
~42 output tk
CONFIDENCE
high
≥$40/mo · ≥100 req