Celeris-1 vs Step 3.7 Flash
Which is better for the money: Celeris-1 by Celeris or Step 3.7 Flash by StepFun? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-08-19.
Trade-off
Verdict
Depends on your budget
Step 3.7 Flash scores 18.5 points higher (30.9 vs 12.4) but costs 2.0× as much to run the benchmark ($320 vs $157; both mid-range class). Pick Step 3.7 Flash when the extra intelligence is worth the premium, otherwise Celeris-1.
| Metric | Celeris-1 | Step 3.7 Flash | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 12.4 | 30.9 | +149.2% Step 3.7 Flash is 2.5× more intelligent |
| Total AA cost (lower is better) | $156.79 | $319.97 | +104.1% Step 3.7 Flash is 2.0× more expensive |
| End-to-end time (lower is better) | 0.95s | 24.06s | +2432.6% Step 3.7 Flash is 2432.6% slower |
| Price per 1M tokens (in / out) | $0.20 / $0.70 | $0.20 / $1.15 | |
| Input modalities | Not reported | Not reported |
Step 3.7 Flash vs Celeris-1: 149.2% more intelligence, 104.1% more cost to run the AA benchmark, and 2432.6% slower end-to-end.
Today's Pareto frontier
Against the frontier
- Celeris-1 is off the frontier: at the same or lower cost, GPT-5.6 Luna (xhigh) scores 50.10 for $95.1.
- Step 3.7 Flash is off the frontier: at the same or lower cost, Gemini 3.7 Flash (medium) scores 53.40 for $278.
Celeris-1Step 3.7 FlashToday's Pareto frontierOpen-weights frontierEvery other model
Matchup
Compare something else
Agents:
Accept: text/markdown or ?format=md returns this page as Markdown. JSON: /api/v1/compare/celeris-1/step-3-7-flash · Markdown: /api/v1/compare/celeris-1/step-3-7-flash?format=md · llms.txt · agent skill