low
No obvious waste on this workload
Token mix looks reasonable for the selected mode. Compare cheaper models on Cost before you scale this up.
Cost800 calls / day
One prompt
$0.008745
Streamed
$0.008745
Thread
$0.0254
Daily total
$7.00
GPT-5.4 is $7.00 / day. Cheapest is GPT-5 Nano at $0.1287 / day.
GPT-5 Nano
OpenAI
$0.1287
GPT-4.1 Nano
OpenAI
$0.1294
Gemini 2.5 Flash-Lite
Gemini
$0.1294
Grok 4.1 Fast
SpaceXAI
$0.1629
GPT-4o Mini
OpenAI
$0.1942
GPT-5.4 Nano
OpenAI
$0.4029
GPT-4.1 Mini
OpenAI
$0.5178
GPT-5 Mini
OpenAI
$0.9316
Grok Build 0.1
SpaceXAI
$0.9424
Gemini 2.5 Flash
Gemini
$1.16
Grok 4.3
SpaceXAI
$1.18
Gemini 3 Flash
Gemini
$1.40
o4-mini
OpenAI
$2.06
o3-mini
OpenAI
$2.06
GPT-5.4 Mini
OpenAI
$2.10
Claude Haiku 4.5
Claude
$2.33
Identify1 findings
Prompt
18
System
0
Data
0
Documents
0
History
0
Output
400
Processing
180
ImproveGPT-5.4
01 · cost
Save ~$6.87 / day
Try GPT-5 Nano for this workload
GPT-5 Nano prices the same tokens at 2% of GPT-5.4.
Switch the selected model on Measure. Keep a flagship only for the hard 10% of calls.
02 · cost
Use Batch APIs for offline work
OpenAI, Anthropic, and Gemini batch endpoints are typically 50% off. SpaceXAI batch is about 20% off.
Route evals, backfills, and overnight jobs to batch. Keep interactive Measure/streamed traffic on the live API.
03 · quality
Route by difficulty, not by habit
A single flagship model for every conversation is the most common overspend pattern.
Classify the prompt first with a nano/flash/haiku model, then escalate only when confidence is low.