Up to four models
DeepSeek-V4.1-Flash vs Step 5 Preview
On the index (GPT-4 = 100), DeepSeek-V4.1-Flash scores 163 and Step 5 Preview 163. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
163147–179
WeightsOpen
ReleasedSep 2026
Price in / out$0.051 / $0.57
Context1.05M
Core tests taken6 of 9
163145–180
WeightsClosed
ReleasedSep 2026
Price in / out$1.0 / $2.7
Context1M
Core tests taken2 of 9
DeepSeek-V4.1-Flash and Step 5 Preview are too close to call (1 point apart, within the uncertainty), ahead on 1 of 2 shared core tests, and costs 7.9× less.
Test by testthird-party results
Every result as a table
| Test | DeepSeek-V4.1-Flash | Step 5 Preview |
|---|---|---|
| Maths | ||
| ProofBench v1.1 | 54.0% | 42.0% |
| Coding | ||
| Terminal-Bench 4.0 | 19.7% | 31.8% |
| Code Migration | 45.6% | 39.0% |
| IOI | 40.3% | 38.1% |
| Vibe Code Bench v1.1 | 84.7% | 70.4% |
| Agents | ||
| BioMysteryBench | 67.8% | 67.4% |
| SRE Bench | 0.8% | 0.4% |
| Terminal-Bench-Science | 0.0% | 2.9% |
| Professional work | ||
| Finance Agent (v2) | 53.5% | 50.7% |
| EMB | 57.2% | 57.1% |
| Legal Research Bench | 41.3% | 40.4% |
| Public Benefits Bench v1.1 | 64.3% | 58.1% |
| Tax Agent Bench | 26.6% | 27.3% |
| Novel problems | ||
| MysteryMechanism | 21.2% | 18.5% |
Tests at least two of these models have taken. Bold rows are the core tests.