Model comparison
DeepSeek V4 Flash (Max) vs Step 3.7 Flash
Head-to-head evidence from 23 shared benchmark results across 6 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Verified leaderboard positions: DeepSeek V4 Flash (Max) #21; Step 3.7 Flash unranked
BenchAlign evidence: DeepSeek V4 Flash (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 Flash (Max) and Step 3.7 Flash share 23 comparable benchmark results. 2 of 8 categories are comparable. 17 results are unique to DeepSeek V4 Flash (Max); 7 to Step 3.7 Flash.
Updated July 16, 2026- Shared results
- 23
- DeepSeek V4 Flash (Max) only
- 17
- Step 3.7 Flash only
- 7
- Comparable categories
- 2 / 8
Pick DeepSeek V4 Flash (Max) if you want the stronger benchmark profile. Step 3.7 Flash only becomes the better choice if agentic is the priority.
Confidence note. This is a partial-evidence comparison with 23 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 Flash (Max) is clearly ahead on the provisional aggregate, 70 to 57. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
DeepSeek V4 Flash (Max)'s sharpest advantage is in coding, where it averages 68.8 against 56.3. The single biggest benchmark swing on the page is SWE-bench Pro, 52.6% 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 $0.14 input / $0.28 output per 1M tokens for DeepSeek V4 Flash (Max). That is roughly 4.1x on output cost alone. DeepSeek V4 Flash (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 Flash (Max) | Δ | Step 3.7 Flash |
|---|---|---|---|
| Coding | DeepSeek V4 Flash (Max)68.8 | Margin← 12.5 | Step 3.7 Flash56.3 |
| Agentic | DeepSeek V4 Flash (Max)63.8 | Margin→ 2.6 | Step 3.7 Flash66.4 |
| Knowledge | DeepSeek V4 Flash (Max)55.6 | MarginNo overlap | Step 3.7 FlashNot measured |
| Math | DeepSeek V4 Flash (Max)94.8 | 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 52.6%B 56.3%Winner: Step 3.7 FlashΔ 3.7SWE-bench Pro: DeepSeek V4 Flash (Max) scored 52.6%; Step 3.7 Flash scored 56.3%. Step 3.7 Flash wins this benchmark. - Source ↗
BrowseComp
AgenticA 73.2%B 75.8%Winner: Step 3.7 FlashΔ 2.6BrowseComp: DeepSeek V4 Flash (Max) scored 73.2%; Step 3.7 Flash scored 75.8%. Step 3.7 Flash wins this benchmark. - Source ↗
Terminal-Bench 2.0
AgenticA 56.9%B 59.5%Winner: Step 3.7 FlashΔ 2.6Terminal-Bench 2.0: DeepSeek V4 Flash (Max) scored 56.9%; 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 Flash (Max) | Step 3.7 Flash | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | DeepSeek V4 Flash (Max)$0.14 input / $0.28 output | Step 3.7 Flash$0.2 input / $1.15 output | DeepSeek V4 Flash (Max) has the lower combined listed price. |
| Generation speedtokens per second | DeepSeek V4 Flash (Max)Not available | Step 3.7 FlashNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | DeepSeek V4 Flash (Max)Not available | Step 3.7 FlashNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | DeepSeek V4 Flash (Max)1M | Step 3.7 Flash256K | DeepSeek V4 Flash (Max) lists the larger context window. |
Benchmark Deep Dive
AgenticStep 3.7 Flash wins13 benchmarks
| Benchmark | DeepSeek V4 Flash (Max) | Step 3.7 Flash | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 56.9% | 59.5% | Step 3.7 Flash leads |
| BrowseCompSource | 73.2% | 75.8% | Step 3.7 Flash leads |
| HLE w/ toolsSource | 45.1% | 47.2% | Step 3.7 Flash leads |
| MCP AtlasSource | 69% | — | Not comparable |
| GDPval-AASource | 1189 | 1017 | DeepSeek V4 Flash (Max) leads |
| ToolathlonSource | 47.8% | 49.5% | Step 3.7 Flash leads |
| AA Agentic IndexSource | 31.1% | 21.5% | DeepSeek V4 Flash (Max) leads |
| τ²-bench resultsSource | 95% | 98.5% | Step 3.7 Flash leads |
| GDPval-AASource | 34.4% | 25.9% | DeepSeek V4 Flash (Max) leads |
| DeepSearchQASource | — | 92.8% | Not comparable |
| Claw-EvalSource | — | 67.1% | Not comparable |
| Gert LabsSource | — | 51.57% | Not comparable |
| APEX-Agents-AASource | — | 14.8% | Not comparable |
CodingDeepSeek V4 Flash (Max) wins9 benchmarks
| Benchmark | DeepSeek V4 Flash (Max) | Step 3.7 Flash | Result |
|---|---|---|---|
| LiveCodeBench Pass@1-COTSource | 91.6% | — | Not comparable |
| CodeforcesSource | 3052.0 | — | Not comparable |
| SWE-bench VerifiedSource | 79% | — | Not comparable |
| SWE-bench ProSource | 52.6% | 56.3% | Step 3.7 Flash leads |
| SWE MultilingualSource | 73.3% | — | Not comparable |
| Terminal-Bench 2.0Source | 56.9% | 59.5% | Step 3.7 Flash leads |
| AA Coding IndexSource | 56.2% | 39.6% | DeepSeek V4 Flash (Max) leads |
| Terminal-Bench HardSource | 35.6% | 35.6% | Tie |
| AA-SciCodeSource | 44.9% | 40.0% | DeepSeek V4 Flash (Max) leads |
Reasoning4 benchmarks
Knowledge12 benchmarks
| Benchmark | DeepSeek V4 Flash (Max) | Step 3.7 Flash | Result |
|---|---|---|---|
| MMLU-ProSource | 86.2% | — | Not comparable |
| SimpleQASource | 34.1% | — | Not comparable |
| Chinese-SimpleQASource | 78.9% | — | Not comparable |
| GPQASource | 88.1% | — | Not comparable |
| GPQA-DSource | 88.1% | — | Not comparable |
| HLESource | 34.8% | — | Not comparable |
| Artificial Analysis Intelligence IndexSource | 40.3% | 30.3% | DeepSeek V4 Flash (Max) leads |
| AA-GPQA DiamondSource | 89.4% | 80.9% | DeepSeek V4 Flash (Max) leads |
| AA-HLESource | 32.1% | 19.9% | DeepSeek V4 Flash (Max) leads |
| AA-Omniscience IndexSource | -22.9% | -37.5% | DeepSeek V4 Flash (Max) leads |
| AA-Omniscience AccuracySource | 37.2% | 25.4% | DeepSeek V4 Flash (Max) leads |
| AA-Omniscience Hallucination RateSource | 95.8% | 84.4% | Step 3.7 Flash leads |
Math4 benchmarks
Multimodal4 benchmarks
Inst. Following1 benchmarks
| Benchmark | DeepSeek V4 Flash (Max) | Step 3.7 Flash | Result |
|---|---|---|---|
| AA-IFBenchSource | 79.2% | 67.3% | DeepSeek V4 Flash (Max) leads |
Frequently Asked Questions (3)
Which is better, DeepSeek V4 Flash (Max) or Step 3.7 Flash?
DeepSeek V4 Flash (Max) is ahead on BenchLM's provisional leaderboard, 70 to 57. The biggest single separator in this matchup is SWE-bench Pro, where the scores are 52.6% and 56.3%.
Which is better for coding, DeepSeek V4 Flash (Max) or Step 3.7 Flash?
DeepSeek V4 Flash (Max) has the edge for coding in this comparison, averaging 68.8 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 Flash (Max) or Step 3.7 Flash?
Step 3.7 Flash has the edge for agentic tasks in this comparison, averaging 66.4 versus 63.8. 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.