DeepSeek V3 (Dec '24) vs DeepSeek V4 Flash 0420 (High)
Which is better for the money: DeepSeek V3 (Dec '24) by DeepSeek or DeepSeek V4 Flash 0420 (High) by DeepSeek? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-10-25.
Trade-off
Verdict
Depends on your budget
DeepSeek V4 Flash 0420 (High) scores 15.9 points higher (24.4 vs 8.5) but costs 4.8× as much to run the benchmark ($121 vs $25.4; mid-range vs budget cost classes). They are not alternatives at one budget: pick DeepSeek V4 Flash 0420 (High) when you need intelligence above 8.5, and DeepSeek V3 (Dec '24) when 8.5 is enough.
| Metric | DeepSeek V3 (Dec '24) | DeepSeek V4 Flash 0420 (High) | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 8.5 | 24.4 | +187.1% DeepSeek V4 Flash 042… is 2.9× more intelligent |
| Total AA cost (lower is better) | $25.37 | $121.20 | +377.7% DeepSeek V4 Flash 042… is 4.8× more expensive |
| End-to-end time (lower is better) | 13.85s | Not reported | — not comparable |
| Price per 1M tokens (in / out) | $0.32 / $0.89 | $0.11 / $0.23 | |
| Input modalities | Text | Text |
DeepSeek V4 Flash 0420 (High) vs DeepSeek V3 (Dec '24): 187.1% more intelligence, 377.7% more cost to run the AA benchmark, and end-to-end time not fully reported.
Today's Pareto frontier
Against the frontier
- DeepSeek V3 (Dec '24) is off the frontier: at the same or lower cost, GPT-6 Luna (Low) scores 21.50 for $10.6.
- DeepSeek V4 Flash 0420 (High) is off the frontier: at the same or lower cost, MiMo-V2.6-Flash scores 37.90 for $110.
DeepSeek V3 (Dec '24)DeepSeek V4 Flash 0420 (High)Today'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-v3/deepseek-v4-flash-0420-high · Markdown: /api/v1/compare/deepseek-v3/deepseek-v4-flash-0420-high?format=md · llms.txt · agent skill