Model comparison
DeepSeek V3.2 vs DeepSeek V4 Pro
Head-to-head evidence from 3 shared benchmark results across 2 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: DeepSeek V3.2 #82 (Supported); DeepSeek V4 Pro #46 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. DeepSeek V3.2 and DeepSeek V4 Pro share 3 comparable benchmark results. 2 of 8 categories are comparable. 16 results are unique to DeepSeek V3.2; 20 to DeepSeek V4 Pro.
Updated July 23, 2026- Shared results
- 3
- DeepSeek V3.2 only
- 16
- DeepSeek V4 Pro only
- 20
- Comparable categories
- 2 / 8
Pick DeepSeek V4 Pro if you want the stronger benchmark profile. DeepSeek V3.2 only becomes the better choice if you want the cheaper token bill.
Confidence note. This is a partial-evidence comparison with 3 shared benchmark results across 2 evidence categories; 2 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
DeepSeek V4 Pro is clearly ahead on the BenchAlign aggregate, 60.66 to 55.4. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
DeepSeek V4 Pro's sharpest advantage is in mathematics, where it averages 31.7 against 17.1.
DeepSeek V4 Pro is also the more expensive model on tokens at $0.43 input / $0.87 output per 1M tokens, versus $0.28 input / $0.42 output per 1M tokens for DeepSeek V3.2. That is roughly 2.1x on output cost alone. DeepSeek V4 Pro gives you the larger context window at 1M, compared with 128K for DeepSeek V3.2.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | DeepSeek V3.2 | Δ | DeepSeek V4 Pro |
|---|---|---|---|
| Math | DeepSeek V3.217.1 | Margin→ 14.6 | DeepSeek V4 Pro31.7 |
| Coding | DeepSeek V3.260.9 | Margin→ 4.4 | DeepSeek V4 Pro65.3 |
| Agentic | DeepSeek V3.2Not measured | MarginNo overlap | DeepSeek V4 Pro59.1 |
| Knowledge | DeepSeek V3.2Not measured | MarginNo overlap | DeepSeek V4 Pro41.3 |
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | DeepSeek V3.2 | DeepSeek V4 Pro | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | DeepSeek V3.2$0.28 input / $0.42 output | DeepSeek V4 Pro$0.435 input / $0.87 output | DeepSeek V3.2 has the lower combined listed price. |
| Generation speedtokens per second | DeepSeek V3.235 tok/s | DeepSeek V4 ProNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | DeepSeek V3.23.75 s | DeepSeek V4 ProNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | DeepSeek V3.2128K | DeepSeek V4 Pro1M | DeepSeek V4 Pro lists the larger context window. |
Benchmark Deep Dive
Agentic8 benchmarks
| Benchmark | DeepSeek V3.2 | DeepSeek V4 Pro | Result |
|---|---|---|---|
| Claw-EvalSource | 40.2% | 59.8% | DeepSeek V4 Pro leads |
| VITA-BenchSource | 18.5% | — | Not comparable |
| τ²-bench resultsSource | 78.9% | — | Not comparable |
| Gert LabsSource | 29.57% | 50.28% | DeepSeek V4 Pro leads |
| Terminal-Bench 2.0Source | — | 59.1% | Not comparable |
| MCP AtlasSource | — | 69.4% | Not comparable |
| ToolathlonSource | — | 46.3% | Not comparable |
| ResearchClawBenchSource | — | 17.1% | Not comparable |
CodingDeepSeek V4 Pro wins7 benchmarks
| Benchmark | DeepSeek V3.2 | DeepSeek V4 Pro | Result |
|---|---|---|---|
| SWE-RebenchSource | 60.9% | — | Not comparable |
| React Native EvalsSource | 71.5% | — | Not comparable |
| AA-SciCodeSource | 38.7% | — | Not comparable |
| SWE-bench VerifiedSource | — | 73.6% | Not comparable |
| SWE-bench ProSource | — | 52.1% | Not comparable |
| SWE MultilingualSource | — | 69.8% | Not comparable |
| Terminal-Bench 2.0Source | — | 59.1% | Not comparable |
Reasoning4 benchmarks
Knowledge12 benchmarks
| Benchmark | DeepSeek V3.2 | DeepSeek V4 Pro | Result |
|---|---|---|---|
| Artificial Analysis Intelligence IndexSource | 24.7% | — | Not comparable |
| AA-GPQA DiamondSource | 75.1% | — | Not comparable |
| AA-HLESource | 10.5% | — | Not comparable |
| AA-Omniscience IndexSource | -46.7% | — | Not comparable |
| AA-Omniscience AccuracySource | 24.2% | — | Not comparable |
| AA-Omniscience Hallucination RateSource | 93.5% | — | Not comparable |
| MMLU-ProSource | — | 82.9% | Not comparable |
| SimpleQASource | — | 45% | Not comparable |
| Chinese-SimpleQASource | — | 75.8% | Not comparable |
| GPQASource | — | 72.9% | Not comparable |
| GPQA-DSource | — | 72.9% | Not comparable |
| HLESource | — | 7.7% | Not comparable |
MathDeepSeek V4 Pro wins6 benchmarks
| Benchmark | DeepSeek V3.2 | DeepSeek V4 Pro | Result |
|---|---|---|---|
| FrontierMath v2 (Tiers 1-3)Source | 22.100% | — | Not comparable |
| FrontierMath v2 (Tier 4)Source | 2.100% | — | Not comparable |
| HMMT Feb 2026Source | — | 31.7% | Not comparable |
| IMOAnswerBenchSource | — | 35.3% | Not comparable |
| ApexSource | — | 0.4% | Not comparable |
| Apex ShortlistSource | — | 9.2% | Not comparable |
Multimodal1 benchmarks
| Benchmark | DeepSeek V3.2 | DeepSeek V4 Pro | Result |
|---|---|---|---|
| Design Arena WebsiteSource | 1204 | 1264 | DeepSeek V4 Pro leads |
Inst. Following1 benchmarks
| Benchmark | DeepSeek V3.2 | DeepSeek V4 Pro | Result |
|---|---|---|---|
| AA-IFBenchSource | 49.0% | — | Not comparable |
Frequently Asked Questions (3)
Which is better, DeepSeek V3.2 or DeepSeek V4 Pro?
DeepSeek V4 Pro is ahead on BenchLM's BenchAlign leaderboard, 60.66 to 55.4.
Which is better for coding, DeepSeek V3.2 or DeepSeek V4 Pro?
DeepSeek V4 Pro has the edge for coding in this comparison, averaging 65.3 versus 60.9. DeepSeek V3.2 stays close enough that the answer can still flip depending on your workload.
Which is better for math, DeepSeek V3.2 or DeepSeek V4 Pro?
DeepSeek V4 Pro has the edge for math in this comparison, averaging 31.7 versus 17.1. DeepSeek V3.2 stays close enough that the answer can still flip depending on your workload.
Related Comparisons
Explore More
Choose a model with this week’s evidence
Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.
One email each week. Unsubscribe anytime.