Deepseek V4 Flash High PreviewvsDeepseek V3.2 Thinking
Deepseek V4 Flash High Preview leads the overall Podium Score by 1.0 points (72.0 vs 71.0).Category wins: Deepseek V4 Flash High Preview 0 — 0 Deepseek V3.2 Thinking.
Deepseek V4 Flash High PreviewDrag slider or hover to compareDeepseek V3.2 Thinking
Quick VerdictAggregated 2026 Benchmark Analysis
Deepseek V4 Flash High Preview is rated higher overall with a Podium Score of 72.0 (vs 71.0 for Deepseek V3.2 Thinking). Both models show closely matched capabilities across domain benchmarks.
Coding & Engineering
Comparable
Inference Speed
Comparable
Cost Efficiency
Comparable
API Price / 1M (in · out)
Deepseek V4 Flash High $0.27/M · $1.1/M — Deepseek V3.2 T $0.27/M · $1.1/M
Choose Deepseek V4 Flash High Preview for general-purpose workloads, and Deepseek V3.2 Thinking for general-purpose workloads.
| Metric | Deepseek V4 Flash High Preview | Deepseek V3.2 Thinking |
|---|---|---|
| Podium Score | 72.0 ▲ | 71.0 |
| Arena Elo | — | — |
| Intelligence Index | — | — |
| Coding | — | — |
| Math | — | — |
| Reasoning | — | — |
| Agentic | — | — |
| Knowledge | — | — |
| Multimodal | — | — |
| Long-context | — | — |
| Output speed | 50 t/s | 50 t/s |
| Output price | $1.1/M | $1.1/M |
| Context window | 128K | 128K |
Scores aggregated from five independent leaderboards. See Methodology. More head-to-heads: Deepseek V4 Flash High vs Claude Mythos · Deepseek V4 Flash High vs Claude 5.5 Sonnet · Deepseek V4 Flash High vs Kimi K3.1 Pro · Deepseek V4 Flash High vs Eleven Multilingual v3