# DeepSeek V4 Pro 0813 vs Step 5 Preview

> Side-by-side benchmark comparison for DeepSeek V4 Pro 0813 and Step 5 Preview across agentic, coding, multimodal, knowledge, reasoning, multilingual, and math tasks.

- Canonical page: https://benchlm.ai/compare/deepseek-v4-pro-0813-vs-step-5-preview
- Last updated: September 21, 2026

- Shared sourced benchmarks: 11
- HTML indexing: indexable
- Ranking lane: BenchAlign v5

## Quick Verdict

Pick Step 5 Preview if you want the stronger benchmark profile. DeepSeek V4 Pro 0813 only makes more sense when its price, context window, or workload-specific category wins matter more than the overall score.

## Summary

- Step 5 Preview leads overall 72 to 63.99.
- The clearest category separation is in reasoning, where the averages are 60.3 for DeepSeek V4 Pro 0813 and 78.5 for Step 5 Preview.
- The biggest single benchmark swing is AutomationBench in Agentic, with scores of 31.8% and 44.0%.
- DeepSeek V4 Pro 0813 is the cheaper option on output tokens, which matters if you expect large responses or heavy interactive use.

## Model Snapshot

| Property | DeepSeek V4 Pro 0813 | Step 5 Preview |
|----------|----------|----------|
| Creator | DeepSeek | StepFun |
| Type | Proprietary | Pending |
| Reasoning | Reasoning | Reasoning |
| Context | 1M | 1M |
| Overall Score | 63.99 | null |
| Benchmarks Covered | 40 | 22 |
| Pricing (input/output) | $0.435 / $0.87 | $1.00 / $2.70 |

## Category Breakdown

### Agentic

- Winner: Coming soon
- DeepSeek V4 Pro 0813 public-lane score: 55.4 (Supported · #36/154)
- Step 5 Preview public-lane score: Coming soon

| Benchmark | DeepSeek V4 Pro 0813 | Step 5 Preview | Winner |
|-----------|-----------|-----------|--------|
| Terminal-Bench 2.0 | 67.9% | Coming soon | Coming soon |
| Terminal-Bench 2.1 | 87.9% | 85.0% | DeepSeek V4 Pro 0813 |
| BrowseComp | 83.4% | 88.7% | Step 5 Preview |
| HLE w/ tools | 60.0% | Coming soon | Coming soon |
| MCP Atlas | 73.6% | 85.6% | Step 5 Preview |
| Toolathlon | 51.8% | Coming soon | Coming soon |
| CyberGym | 83.3% | 84.7% | Step 5 Preview |
| Toolathlon-Verified | 74.1% | 74.1% | Tie |
| Agents' Last Exam | 25.7% | 29.5% | Step 5 Preview |
| AutomationBench | 31.8% | 44.0% | Step 5 Preview |
| Terminal-Bench 2.1 (Vals) | 54.7% | Coming soon | Coming soon |
| Terminal-Bench 4.0 | Coming soon | 33.30% | Coming soon |
| JobBench | Coming soon | 59.0% | Coming soon |
| APEX-Agents | Coming soon | 37.8% | Coming soon |
| DRACO | Coming soon | 83.3% | Coming soon |

### Coding

- Winner: Coming soon
- DeepSeek V4 Pro 0813 public-lane score: 52.5 (Supported · #50/156)
- Step 5 Preview public-lane score: Coming soon

| Benchmark | DeepSeek V4 Pro 0813 | Step 5 Preview | Winner |
|-----------|-----------|-----------|--------|
| LiveCodeBench Pass@1-COT | 93.5% | Coming soon | Coming soon |
| Codeforces | 3206.0 | Coming soon | Coming soon |
| SWE-bench Verified | 80.6% | Coming soon | Coming soon |
| SWE-bench Pro | 55.4% | Coming soon | Coming soon |
| SWE Multilingual | 76.2% | Coming soon | Coming soon |
| Terminal-Bench 2.0 | 67.9% | Coming soon | Coming soon |
| Vibe Code Bench | 49.93% | Coming soon | Coming soon |
| Terminal-Bench 2.1 | 87.9% | 85.0% | DeepSeek V4 Pro 0813 |
| NL2Repo | 61.5% | Coming soon | Coming soon |
| DeepSWE | 62.7% | 67.7% | Step 5 Preview |
| DSBench-FullStack | 71.1% | Coming soon | Coming soon |
| DSBench-Hard | 67.2% | Coming soon | Coming soon |
| OpenHarmony Bench | 59.0% | Coming soon | Coming soon |
| LiveCodeBench (Vals) | 87.5% | Coming soon | Coming soon |
| SWE-bench (Vals) | 96.4% | Coming soon | Coming soon |
| SciCode | Coming soon | 58.9% | Coming soon |
| ProgramBench | Coming soon | 80.5% | Coming soon |
| sweMarathon | Coming soon | 72.7% | Coming soon |
| MLS-Bench Lite | Coming soon | 40.5% | Coming soon |

### Reasoning

- Winner: Coming soon
- DeepSeek V4 Pro 0813 public-lane score: 60.3 (#14/19)
- Step 5 Preview public-lane score: 78.5 (Unranked · 2 rankable rows)

| Benchmark | DeepSeek V4 Pro 0813 | Step 5 Preview | Winner |
|-----------|-----------|-----------|--------|
| MRCR 1M | 83.5% | Coming soon | Coming soon |
| CorpusQA 1M | 62.0% | Coming soon | Coming soon |
| CritPt | Coming soon | 20.9% | Coming soon |

### Multimodal & Grounded

- Winner: Coming soon
- DeepSeek V4 Pro 0813 public-lane score: Coming soon
- Step 5 Preview public-lane score: 61.2 (#28/49)

| Benchmark | DeepSeek V4 Pro 0813 | Step 5 Preview | Winner |
|-----------|-----------|-----------|--------|
| OfficeQA Pro | Coming soon | 60.3% | Coming soon |
| MMMU-Pro | Coming soon | 76% | Coming soon |

### Knowledge

- Winner: Coming soon
- DeepSeek V4 Pro 0813 public-lane score: 58.2 (Estimated · #42/186)
- Step 5 Preview public-lane score: Coming soon

| Benchmark | DeepSeek V4 Pro 0813 | Step 5 Preview | Winner |
|-----------|-----------|-----------|--------|
| MMLU-Pro | 87.5% | Coming soon | Coming soon |
| SimpleQA | 57.9% | Coming soon | Coming soon |
| Chinese-SimpleQA | 84.4% | Coming soon | Coming soon |
| GPQA | 90.1% | Coming soon | Coming soon |
| GPQA-D | 90.1% | 93.5% | Step 5 Preview |
| HLE | 42.7% | 46.5% | Step 5 Preview |
| GPQA Diamond (Vals) | 92.4% | Coming soon | Coming soon |
| MMLU-Pro (Vals) | 87.0% | Coming soon | Coming soon |

### Mathematics

- Winner: Coming soon
- DeepSeek V4 Pro 0813 public-lane score: 80.2 (Unranked · 4 rankable rows)
- Step 5 Preview public-lane score: Coming soon

| Benchmark | DeepSeek V4 Pro 0813 | Step 5 Preview | Winner |
|-----------|-----------|-----------|--------|
| HMMT Feb 2026 | 95.2% | Coming soon | Coming soon |
| IMOAnswerBench | 89.8% | Coming soon | Coming soon |
| Apex | 38.3% | Coming soon | Coming soon |
| Apex Shortlist | 90.2% | Coming soon | Coming soon |

## FAQ

### Which is better overall, DeepSeek V4 Pro 0813 or Step 5 Preview?

Step 5 Preview is ahead overall on BenchLM right now.

### Where is the biggest gap between DeepSeek V4 Pro 0813 and Step 5 Preview?

The widest category gap is in reasoning, where the averages are 60.3 for DeepSeek V4 Pro 0813 and 78.5 for Step 5 Preview.

### How many benchmarks does DeepSeek V4 Pro 0813 cover on BenchLM?

DeepSeek V4 Pro 0813 currently has 40 sourced benchmark scores on BenchLM.

### How many benchmarks does Step 5 Preview cover on BenchLM?

Step 5 Preview currently has 22 sourced benchmark scores on BenchLM.

## Related Comparisons

- [DeepSeek V4 Pro 0813 vs Claude Fable 5.1](/compare/claude-fable-5-1-vs-deepseek-v4-pro-0813)
- [Step 5 Preview vs Claude Fable 5.1](/compare/claude-fable-5-1-vs-step-5-preview)
- [DeepSeek V4 Pro 0813 vs GPT-6 Astra](/compare/deepseek-v4-pro-0813-vs-gpt-6-astra)
- [Step 5 Preview vs GPT-6 Astra](/compare/gpt-6-astra-vs-step-5-preview)

## Explore More

- [DeepSeek V4 Pro 0813 profile](/models/deepseek-v4-pro-0813)
- [Step 5 Preview profile](/models/step-5-preview)
- [Compare Pricing](/llm-pricing)
- [Alternative Finder](/tools/alternative-finder)
- [LLM Selector](/tools/llm-selector)
- [Overall Rankings](/best/overall)