Model comparison
DeepSeek V4 Pro (Max) vs Step 3.7 Flash
Head-to-head evidence from 24 shared benchmark results across 6 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Verified leaderboard positions: DeepSeek V4 Pro (Max) #9; Step 3.7 Flash unranked
BenchAlign evidence: DeepSeek V4 Pro (Max) not scored; Step 3.7 Flash estimated. Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. DeepSeek V4 Pro (Max) and Step 3.7 Flash share 24 comparable benchmark results. 2 of 8 categories are comparable. 18 results are unique to DeepSeek V4 Pro (Max); 6 to Step 3.7 Flash.
Updated July 16, 2026- Shared results
- 24
- DeepSeek V4 Pro (Max) only
- 18
- Step 3.7 Flash only
- 6
- Comparable categories
- 2 / 8
Pick DeepSeek V4 Pro (Max) if you want the stronger benchmark profile. Step 3.7 Flash only becomes the better choice if its workflow or ecosystem matters more than the raw scoreboard.
Confidence note. This is a partial-evidence comparison with 24 shared benchmark results across 6 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 (Max) is clearly ahead on the provisional aggregate, 76 to 57. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
DeepSeek V4 Pro (Max)'s sharpest advantage is in coding, where it averages 70.9 against 56.3. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 67.9% to 59.5%.
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 (Max). DeepSeek V4 Pro (Max) 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 (Max) | Δ | Step 3.7 Flash |
|---|---|---|---|
| Coding | DeepSeek V4 Pro (Max)70.9 | Margin← 14.6 | Step 3.7 Flash56.3 |
| Agentic | DeepSeek V4 Pro (Max)74.5 | Margin← 8.1 | Step 3.7 Flash66.4 |
| Knowledge | DeepSeek V4 Pro (Max)60.4 | MarginNo overlap | Step 3.7 FlashNot measured |
| Math | DeepSeek V4 Pro (Max)95.2 | MarginNo overlap | Step 3.7 FlashNot measured |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
Terminal-Bench 2.0
AgenticA 67.9%B 59.5%Winner: DeepSeek V4 Pro (Max)Δ 8.4Terminal-Bench 2.0: DeepSeek V4 Pro (Max) scored 67.9%; Step 3.7 Flash scored 59.5%. DeepSeek V4 Pro (Max) wins this benchmark. - Source ↗
BrowseComp
AgenticA 83.4%B 75.8%Winner: DeepSeek V4 Pro (Max)Δ 7.6BrowseComp: DeepSeek V4 Pro (Max) scored 83.4%; Step 3.7 Flash scored 75.8%. DeepSeek V4 Pro (Max) wins this benchmark. - Source ↗
SWE-bench Pro
CodingA 55.4%B 56.3%Winner: Step 3.7 FlashΔ 0.9SWE-bench Pro: DeepSeek V4 Pro (Max) scored 55.4%; Step 3.7 Flash scored 56.3%. 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 (Max) | Step 3.7 Flash | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | DeepSeek V4 Pro (Max)$0.435 input / $0.87 output | Step 3.7 Flash$0.2 input / $1.15 output | DeepSeek V4 Pro (Max) has the lower combined listed price. |
| Generation speedtokens per second | DeepSeek V4 Pro (Max)Not available | Step 3.7 FlashNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | DeepSeek V4 Pro (Max)Not available | Step 3.7 FlashNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | DeepSeek V4 Pro (Max)1M | Step 3.7 Flash256K | DeepSeek V4 Pro (Max) lists the larger context window. |
Benchmark Deep Dive
AgenticDeepSeek V4 Pro (Max) wins13 benchmarks
| Benchmark | DeepSeek V4 Pro (Max) | Step 3.7 Flash | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 67.9% | 59.5% | DeepSeek V4 Pro (Max) leads |
| BrowseCompSource | 83.4% | 75.8% | DeepSeek V4 Pro (Max) leads |
| HLE w/ toolsSource | 48.2% | 47.2% | DeepSeek V4 Pro (Max) leads |
| MCP AtlasSource | 73.6% | — | Not comparable |
| GDPval-AASource | 1307 | 1017 | DeepSeek V4 Pro (Max) leads |
| ToolathlonSource | 51.8% | 49.5% | DeepSeek V4 Pro (Max) leads |
| AA Agentic IndexSource | 36.4% | 21.5% | DeepSeek V4 Pro (Max) leads |
| APEX-Agents-AASource | 24.3% | 14.8% | DeepSeek V4 Pro (Max) leads |
| τ²-bench resultsSource | 96.2% | 98.5% | Step 3.7 Flash leads |
| GDPval-AASource | 40.4% | 25.9% | DeepSeek V4 Pro (Max) leads |
| DeepSearchQASource | — | 92.8% | Not comparable |
| Claw-EvalSource | — | 67.1% | Not comparable |
| Gert LabsSource | — | 51.57% | Not comparable |
CodingDeepSeek V4 Pro (Max) wins10 benchmarks
| Benchmark | DeepSeek V4 Pro (Max) | Step 3.7 Flash | Result |
|---|---|---|---|
| LiveCodeBench Pass@1-COTSource | 93.5% | — | Not comparable |
| CodeforcesSource | 3206.0 | — | Not comparable |
| SWE-bench VerifiedSource | 80.6% | — | Not comparable |
| SWE-bench ProSource | 55.4% | 56.3% | Step 3.7 Flash leads |
| SWE MultilingualSource | 76.2% | — | Not comparable |
| Terminal-Bench 2.0Source | 67.9% | 59.5% | DeepSeek V4 Pro (Max) leads |
| Vibe Code BenchSource | 49.93% | — | Not comparable |
| AA Coding IndexSource | 59.4% | 39.6% | DeepSeek V4 Pro (Max) leads |
| Terminal-Bench HardSource | 46.2% | 35.6% | DeepSeek V4 Pro (Max) leads |
| AA-SciCodeSource | 50.0% | 40.0% | DeepSeek V4 Pro (Max) leads |
Reasoning4 benchmarks
Knowledge12 benchmarks
| Benchmark | DeepSeek V4 Pro (Max) | Step 3.7 Flash | Result |
|---|---|---|---|
| MMLU-ProSource | 87.5% | — | Not comparable |
| SimpleQASource | 57.9% | — | Not comparable |
| Chinese-SimpleQASource | 84.4% | — | Not comparable |
| GPQASource | 90.1% | — | Not comparable |
| GPQA-DSource | 90.1% | — | Not comparable |
| HLESource | 37.7% | — | Not comparable |
| Artificial Analysis Intelligence IndexSource | 44.3% | 30.3% | DeepSeek V4 Pro (Max) leads |
| AA-GPQA DiamondSource | 88.8% | 80.9% | DeepSeek V4 Pro (Max) leads |
| AA-HLESource | 35.9% | 19.9% | DeepSeek V4 Pro (Max) leads |
| AA-Omniscience IndexSource | -10.0% | -37.5% | DeepSeek V4 Pro (Max) leads |
| AA-Omniscience AccuracySource | 43.3% | 25.4% | DeepSeek V4 Pro (Max) leads |
| AA-Omniscience Hallucination RateSource | 94.0% | 84.4% | Step 3.7 Flash leads |
Math4 benchmarks
Multimodal4 benchmarks
Inst. Following1 benchmarks
| Benchmark | DeepSeek V4 Pro (Max) | Step 3.7 Flash | Result |
|---|---|---|---|
| AA-IFBenchSource | 76.5% | 67.3% | DeepSeek V4 Pro (Max) leads |
Frequently Asked Questions (3)
Which is better, DeepSeek V4 Pro (Max) or Step 3.7 Flash?
DeepSeek V4 Pro (Max) is ahead on BenchLM's provisional leaderboard, 76 to 57. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 67.9% and 59.5%.
Which is better for coding, DeepSeek V4 Pro (Max) or Step 3.7 Flash?
DeepSeek V4 Pro (Max) has the edge for coding in this comparison, averaging 70.9 versus 56.3. Inside this category, AA Coding Index is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, DeepSeek V4 Pro (Max) or Step 3.7 Flash?
DeepSeek V4 Pro (Max) has the edge for agentic tasks in this comparison, averaging 74.5 versus 66.4. Inside this category, GDPval-AA 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.