Model comparison
DeepSeek V4 Pro vs Step 3.7 Flash
Head-to-head evidence from 7 shared benchmark results across 3 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
BenchAlign evidence: DeepSeek V4 Pro supported; Step 3.7 Flash estimated. Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. DeepSeek V4 Pro and Step 3.7 Flash share 7 comparable benchmark results. 2 of 8 categories are comparable. 17 results are unique to DeepSeek V4 Pro; 23 to Step 3.7 Flash.
Updated July 16, 2026- Shared results
- 7
- DeepSeek V4 Pro only
- 17
- Step 3.7 Flash only
- 23
- Comparable categories
- 2 / 8
Pick DeepSeek V4 Pro if you want the stronger benchmark profile. Step 3.7 Flash only becomes the better choice if agentic is the priority or you want the stronger reasoning-first profile.
Confidence note. This is a partial-evidence comparison with 7 shared benchmark results across 3 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 provisional aggregate, 64 to 57. 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 coding, where it averages 65.3 against 56.3. The single biggest benchmark swing on the page is SWE-bench Pro, 52.1% to 56.3%. Step 3.7 Flash does hit back in agentic, so the answer changes if that is the part of the workload you care about most.
Step 3.7 Flash is also the more expensive model on tokens at $0.20 input / $1.15 output per 1M tokens, versus $0.43 input / $0.87 output per 1M tokens for DeepSeek V4 Pro. Step 3.7 Flash is the reasoning model in the pair, while DeepSeek V4 Pro is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. DeepSeek V4 Pro gives you the larger context window at 1M, compared with 256K for Step 3.7 Flash.
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 V4 Pro | Δ | Step 3.7 Flash |
|---|---|---|---|
| Coding | DeepSeek V4 Pro65.3 | Margin← 9.0 | Step 3.7 Flash56.3 |
| Agentic | DeepSeek V4 Pro59.1 | Margin→ 7.3 | Step 3.7 Flash66.4 |
| Knowledge | DeepSeek V4 Pro41.7 | MarginNo overlap | Step 3.7 FlashNot measured |
| Math | DeepSeek V4 Pro31.7 | MarginNo overlap | Step 3.7 FlashNot measured |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
SWE-bench Pro
CodingA 52.1%B 56.3%Winner: Step 3.7 FlashΔ 4.2SWE-bench Pro: DeepSeek V4 Pro scored 52.1%; Step 3.7 Flash scored 56.3%. Step 3.7 Flash wins this benchmark. - Source ↗
Terminal-Bench 2.0
AgenticA 59.1%B 59.5%Winner: Step 3.7 FlashΔ 0.4Terminal-Bench 2.0: DeepSeek V4 Pro scored 59.1%; Step 3.7 Flash scored 59.5%. Step 3.7 Flash wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | DeepSeek V4 Pro | Step 3.7 Flash | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | DeepSeek V4 Pro$0.435 input / $0.87 output | Step 3.7 Flash$0.2 input / $1.15 output | DeepSeek V4 Pro has the lower combined listed price. |
| Generation speedtokens per second | DeepSeek V4 ProNot available | Step 3.7 FlashNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | DeepSeek V4 ProNot available | Step 3.7 FlashNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | DeepSeek V4 Pro1M | Step 3.7 Flash256K | DeepSeek V4 Pro lists the larger context window. |
Benchmark Deep Dive
AgenticStep 3.7 Flash wins14 benchmarks
| Benchmark | DeepSeek V4 Pro | Step 3.7 Flash | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 59.1% | 59.5% | Step 3.7 Flash leads |
| MCP AtlasSource | 69.4% | — | Not comparable |
| ToolathlonSource | 46.3% | 49.5% | Step 3.7 Flash leads |
| Claw-EvalSource | 59.8% | 67.1% | Step 3.7 Flash leads |
| Gert LabsSource | 50.28% | 51.57% | Step 3.7 Flash leads |
| ResearchClawBenchSource | 17.1% | — | Not comparable |
| BrowseCompSource | — | 75.8% | Not comparable |
| DeepSearchQASource | — | 92.8% | Not comparable |
| GDPval-AASource | — | 25.9% | Not comparable |
| HLE w/ toolsSource | — | 47.2% | Not comparable |
| AA Agentic IndexSource | — | 21.5% | Not comparable |
| τ²-bench resultsSource | — | 98.5% | Not comparable |
| GDPval-AASource | — | 1017 | Not comparable |
| APEX-Agents-AASource | — | 14.8% | Not comparable |
CodingDeepSeek V4 Pro wins8 benchmarks
| Benchmark | DeepSeek V4 Pro | Step 3.7 Flash | Result |
|---|---|---|---|
| LiveCodeBench Pass@1-COTSource | 56.8% | — | Not comparable |
| SWE-bench VerifiedSource | 73.6% | — | Not comparable |
| SWE-bench ProSource | 52.1% | 56.3% | Step 3.7 Flash leads |
| SWE MultilingualSource | 69.8% | — | Not comparable |
| Terminal-Bench 2.0Source | 59.1% | 59.5% | Step 3.7 Flash leads |
| AA Coding IndexSource | — | 39.6% | Not comparable |
| Terminal-Bench HardSource | — | 35.6% | Not comparable |
| AA-SciCodeSource | — | 40.0% | Not comparable |
Reasoning4 benchmarks
Knowledge12 benchmarks
| Benchmark | DeepSeek V4 Pro | Step 3.7 Flash | Result |
|---|---|---|---|
| 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 |
| Artificial Analysis Intelligence IndexSource | — | 30.3% | Not comparable |
| AA-GPQA DiamondSource | — | 80.9% | Not comparable |
| AA-HLESource | — | 19.9% | Not comparable |
| AA-Omniscience IndexSource | — | -37.5% | Not comparable |
| AA-Omniscience AccuracySource | — | 25.4% | Not comparable |
| AA-Omniscience Hallucination RateSource | — | 84.4% | Not comparable |
Math4 benchmarks
Multimodal4 benchmarks
Inst. Following1 benchmarks
| Benchmark | DeepSeek V4 Pro | Step 3.7 Flash | Result |
|---|---|---|---|
| AA-IFBenchSource | — | 67.3% | Not comparable |
Frequently Asked Questions (3)
Which is better, DeepSeek V4 Pro or Step 3.7 Flash?
DeepSeek V4 Pro is ahead on BenchLM's provisional leaderboard, 64 to 57. The biggest single separator in this matchup is SWE-bench Pro, where the scores are 52.1% and 56.3%.
Which is better for coding, DeepSeek V4 Pro or Step 3.7 Flash?
DeepSeek V4 Pro has the edge for coding in this comparison, averaging 65.3 versus 56.3. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, DeepSeek V4 Pro or Step 3.7 Flash?
Step 3.7 Flash has the edge for agentic tasks in this comparison, averaging 66.4 versus 59.1. Inside this category, Claw-Eval is the benchmark that creates the most daylight between them.
Related Comparisons
Explore More
The AI models change fast. We track them for you.
A weekly brief for engineers and researchers covering new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.