Beam 1.0 vs Step 5 Preview
Which is better for the money: Beam 1.0 by Reflection or Step 5 Preview by StepFun? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-10-17.
Trade-off
Verdict
Depends on your budget
Step 5 Preview scores 13.4 points higher (43.7 vs 30.3) but costs 1.6× as much to run the benchmark ($1190 vs $740; both premium class). Pick Step 5 Preview when the extra intelligence is worth the premium, otherwise Beam 1.0.
| Metric | Beam 1.0 | Step 5 Preview | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 30.3 | 43.7 | +44.2% Step 5 Preview is 44.2% more intelligent |
| Total AA cost (lower is better) | $739.78 | $1,190.26 | +60.9% Step 5 Preview is 60.9% more expensive |
| End-to-end time (lower is better) | 13.31s | 29.86s | +124.3% Step 5 Preview is 124.3% slower |
| Price per 1M tokens (in / out) | $0.70 / $2.10 | $1.00 / $2.70 | |
| Input modalities | Text | Text, Image |
Step 5 Preview vs Beam 1.0: 44.2% more intelligence, 60.9% more cost to run the AA benchmark, and 124.3% slower end-to-end.
Today's Pareto frontier
Against the frontier
- Beam 1.0 is off the frontier: at the same or lower cost, GPT-6.1 Sol (Xhigh) scores 51.00 for $662.
- Step 5 Preview is off the frontier: at the same or lower cost, GPT-6.1 Sol (Max) scores 51.80 for $1082.
Beam 1.0Step 5 PreviewToday'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/reflection-beam/step-5 · Markdown: /api/v1/compare/reflection-beam/step-5?format=md · llms.txt · agent skill