DeepSeek V4.1 Flash 1017 (Non-reasoning) vs Step 5 Preview
Which is better for the money: DeepSeek V4.1 Flash 1017 (Non-reasoning) by DeepSeek or Step 5 Preview by StepFun? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-11-01.
Trade-off
Verdict
Depends on your budget
Step 5 Preview scores 10.0 points higher (28.5 vs 18.5) but costs 5.3× as much to run the benchmark ($1644 vs $310; premium vs mid-range cost classes). They are not alternatives at one budget: pick Step 5 Preview when you need intelligence above 18.5, and DeepSeek V4.1 Flash 1017 (Non-reasoning) when 18.5 is enough.
| Metric | DeepSeek V4.1 Flash 1017 (Non-reasoning) | Step 5 Preview | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 18.5 | 28.5 | +54.1% Step 5 Preview is 54.1% more intelligent |
| Total AA cost (lower is better) | $310.24 | $1,643.75 | +429.8% Step 5 Preview is 5.3× more expensive |
| End-to-end time (lower is better) | 3.92s | 18.82s | +380.1% Step 5 Preview is 380.1% slower |
| Price per 1M tokens (in / out) | $0.30 / $1.20 | $1.00 / $2.70 | |
| Input modalities | Text, Image | Text, Image |
Step 5 Preview vs DeepSeek V4.1 Flash 1017 (Non-reasoning): 54.1% more intelligence, 429.8% more cost to run the AA benchmark, and 380.1% slower end-to-end.
Today's Pareto frontier
Against the frontier
- DeepSeek V4.1 Flash 1017 (Non-reasoning) is off the frontier: at the same or lower cost, GPT-6.1 Terra (High) scores 42.30 for $267.
- Step 5 Preview is off the frontier: at the same or lower cost, GPT-6.1 Sol (Max) scores 54.50 for $1124.
DeepSeek V4.1 Flash 1017 (Non-reasoning)Step 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/deepseek-v4-1-flash-1017-non-reasoning/step-5 · Markdown: /api/v1/compare/deepseek-v4-1-flash-1017-non-reasoning/step-5?format=md · llms.txt · agent skill