Jev next to an LLM
One request, sent to Jev and your pick of 8 OpenAI models at the same moment. Both answers, both latencies, both costs. Each network gets 5 free comparisons a day.
Same question, three models
average latency · drawn to scalejev-1.13.0111 ms
gpt-5.4-mini1.4 s
gpt-5.5 · reasoning11.1 s
TypeSafe’s consistency cookbook, 2026-09-11, 15 runs: Jev averaged 111 ms; the others are its 12.7× and 100.2× slowdowns applied to that. Theirs, not ours — run your own below.
How we measure
1. Same moment
Both calls leave one server function together.2. Own stopwatch
Each side times its own call, body included.3. Real model id
Both sides report the model id that answered. The LLM uses Structured Outputs, no retries, its fastest reasoning setting.4. List-price cost
Tokens × the prices below. Jev’s output is free.
What one run can’t tell you
Accuracy
Agreeing isn’t being right.
Calibration
That takes many runs.
A trend
One run, one region, network included.
LLM confidence
A number it writes isn’t a distribution.
Other people’s numbersshow
- TypeSafe: 193.6× faster and 444.6× cheaper on their own System One workflows; 40–200× faster as the real-world range, and 70–500 ms end to end, per the launch post.
- Community: the jev-test harness found Jev about 3× faster than GPT-5.6 Terra and 7.5× faster than Claude Opus 5, at about 35× and 190× lower cost, on 17 easy cases run once.
- TypeSafe’s consistency cookbook (2026-09-11, 15 runs): gpt-5.4-mini was 12.7× slower and 25.6× costlier; gpt-5.5 with reasoning 100.2× slower and 778.9× costlier. Jev averaged 111 ms.
Theirs, not ours, and measured under their conditions.
Prices used · per 1M tokens
OpenAI pricing, checked 2026-09-22 ↗| Model | Input | Output | Note |
|---|---|---|---|
| jev-1.13.0 | $0.042 | free | TypeSafe, checked 2026-09-21 |
| gpt-5.6-sol | $4 | $20 | Large |
| gpt-5.6-terra | $2 | $12 | Balanced |
| gpt-5.6-luna | $0.2 | $1.2 | Small and fast |
| gpt-5.4-mini | $0.75 | $4.5 | Small |
| gpt-5.4-nano | $0.2 | $1.25 | Smallest 5.4 |
| o4-mini | $1.1 | $4.4 | Reasoning |
| gpt-4.1-mini | $0.4 | $1.6 | No reasoning |
| gpt-4o-mini | $0.15 | $0.6 | No reasoning, cheapest |
Comparisons run from local.