Data points: 46
Model compare
The readout for DeepSeek V4 Pro and Step3, before the detailed comparison sheet.
Weighted outcome: DeepSeek V4 Pro. Benchmark capability categories carry 80%, while price, API performance, and availability carry 20%.
Decision read
DeepSeek V4 Pro
DeepSeek V4 Pro has the higher weighted result; Model A / B score 100 to 0.
Evidence depth
46 data points
Includes 0 benchmark rows, 0 audit samples, and 6 provider examples.
Selection signal
Start with DeepSeek V4 Pro
The charts below split 6 high-signal samples across speed, scores, and audit health.
Switch either side of this report to compare another model with the same LMSpeed data pipeline.
Select a different model to open a new comparison URL.
This report only uses LMSpeed data for DeepSeek V4 Pro and Step3: pricing, speed aggregates, third-party benchmark scores, and shared provider samples.
| Model compare | DeepSeek V4 Pro | Step3 |
|---|---|---|
| Overall leader | Leading | Contender |
| Weighted overall score | 100.0 pts | 0.0 pts |
| Benchmark category leads | 0 categories | 0 categories |
| Operational advantages | Cheapest input price, Free providers, Provider coverage | No data |
| Context window | 1.0M tokens | No data |
| Max output | 384K tokens | No data |
| Modalities | Input Text Output Text | No data |
| Features |
The overall result weights benchmark capability categories at 80% and price, API speed/latency, and availability at 20%. Recent test volume does not affect the winner, and missing benchmark categories are excluded.
| Model compare | DeepSeek V4 Pro | Step3 |
|---|---|---|
| Developer | DeepSeek | No data |
| Released | Apr 2026 | No data |
| Parameters | No data | No data |
| Tokenizer | DeepSeek | No data |
| Knowledge cutoff | No data | No data |
| OpenRouter ID | deepseek/deepseek-v4-pro | No data |
| References | No data | No data |
This report only uses LMSpeed data for DeepSeek V4 Pro and Step3: pricing, speed aggregates, third-party benchmark scores, and shared provider samples.
DeepSeek V4 Pro
DeepSeek V4 Pro has these operational advantages: Cheapest input price, Free providers, Provider coverage.
Step3
Step3 does not clearly lead in the benchmark or operational dimensions shared by both models.
Third-party benchmark profile synced into LMSpeed; only metrics available for both models are shown.
Compare benchmark category scores on a 0-100 scale. Select a category to inspect the gap.
Avg. score
DeepSeek V4 Pro
49.2
Avg. score
Step3
-
DeepSeek V4 Pro
DeepSeek V4 Pro
DeepSeek V4 Pro
DeepSeek V4 Pro
DeepSeek V4 Pro
No data
No data
DeepSeek V4 Pro
Metric-level scores with benchmark source, rank depth, confidence, error, and evaluation date where available.
No shared professional benchmark scores are available yet.
Latest completed audits from shared providers, with four safety and integrity score groups plus report links.
| Provider | DeepSeek V4 Pro | Step3 |
|---|---|---|
| No completed audits are available from shared providers yet. | ||
Speed aggregates and input/output pricing share each provider row for real API selection and migration cost checks.
| Provider | DeepSeek V4 Pro | Step3 |
|---|---|---|
5 tests | DeepSeek V4 Pro speed / latency 81 tok/s / 5166ms input / output No data | Step3 speed / latency N/A / N/A input / output No data |
5 tests | DeepSeek V4 Pro speed / latency 48 tok/s / 4105ms input / output No data | Step3 speed / latency N/A / N/A input / output No data |
0 tests | DeepSeek V4 Pro speed / latency N/A / N/A input / output No data | Step3 speed / latency N/A / N/A input / output No data |
0 tests | DeepSeek V4 Pro speed / latency N/A / N/A input / output No data | Step3 speed / latency N/A / N/A input / output No data |
0 tests | DeepSeek V4 Pro speed / latency N/A / N/A input / output No data | Step3 speed / latency N/A / N/A input / output No data |
DeepSeek V4 Pro deepseek-v4-pro speed / latency No data input / output $3.82/M/$7.64/M | Step3 stepfun-ai/step3 speed / latency No data input / output $6.36/M |
Weighted outcome: DeepSeek V4 Pro. Benchmark capability categories carry 80%, while price, API performance, and availability carry 20%.
Text inputText outputTool callingStructured outputsJSON modeReasoning |
| None listed |
