Model comparison
GPT-4.1 vs Step 3
Step 3 is the stronger model overall, scoring 39.9 to 35.7 on the Noometry Index.
Last verified . 17 shared benchmarks.
StepFunSummary
- They share 17 benchmarks with published results for both. GPT-4.1 scores higher in 5 categories and Step 3 in 4 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Step 3 leads 27.6 to 11.7.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 52.3% for GPT-4.1 and 62.3% for Step 3.
- Step 3 has downloadable open weights; the other is API-only.
Side by side
| GPT-4.1 | Step 3 | |
|---|---|---|
| Provider | OpenAI | StepFun |
| Noometry Index | 35.7 | 39.9 |
| Released | 2025-04-14 | — |
| Weights | Proprietary | Open |
| Context window | 1.05M | — |
| Max output | 33K | — |
| Input $ / M tokens | $2 | — |
| Output $ / M tokens | $8 | — |
| Results tracked | 52 | 17 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Step 3 leads
GPT-4.1: 34.2 (#228), Step 3: 39.5 (#144)
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| LMArena Coding | 1391 | 1367 |
| SWE-bench Verified | 48.5% | — |
| SWE-bench Verified (bash only) | 39.6% | — |
| Aider Polyglot | 52.4% | — |
| WeirdML | 39% | — |
| CadEval | 42% | — |
| ALE-Bench | 558.1 | — |
Agentic & Tool Use Not comparable
GPT-4.1: 34.4 (#40), Step 3: —
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| Berkeley Function Calling Leaderboard | 54% | — |
Reasoning Step 3 leads
GPT-4.1: 11.7 (#334), Step 3: 27.6 (#102)
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| Kagi LLM Benchmark | 52.3% | 62.3% |
| LMArena Hard Prompts | 1384 | 1355 |
| ARC-AGI-2 | 0.4% | — |
| SimpleBench | 27% | — |
| ARC-AGI-1 | 5.5% | — |
| Chess Puzzles | 6% | — |
| EnigmaEval | 2.2% | — |
| DTBench | 68.3% | — |
| LMCA | 25.6% | — |
| Epoch Capabilities Index | 136.79 | — |
| ForecastBench | 61.5 | — |
Math Step 3 leads
GPT-4.1: 21.4 (#269), Step 3: 34.9 (#148)
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| LMArena Math | 1369 | 1366 |
| FrontierMath (Tiers 1-3) | 6% | — |
| OTIS Mock AIME 2024-2025 | 38.3% | — |
| Omni-MATH | 47.1% | — |
| MATH Level 5 | 83% | — |
| FrontierMath (Feb 2025 set) | 5.5% | — |
| FrontierMath Tier 4 (v1) | 0% | — |
Knowledge Too close to call
GPT-4.1: 37.0 (#152), Step 3: 36.2 (#161)
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| LMArena Expert | 1364 | 1333 |
| GPQA Diamond | 66.9% | — |
| Humanity's Last Exam | 5.4% | — |
| SimpleQA Verified | 31.1% | — |
| MMLU-Pro | 81.1% | — |
| Vectara Hallucination Rate | 5.6% | — |
| GPQA (HELM) | 65.9% | — |
Multimodal GPT-4.1 leads
GPT-4.1: 37.5 (#66), Step 3: 34.9 (#81)
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| LMArena Vision | 1210 | 1176 |
| GeoBench | 72% | — |
Multilingual GPT-4.1 leads
GPT-4.1: 49.6 (#128), Step 3: 46.5 (#154)
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| LMArena Non-English | 1370 | 1327 |
| LMArena Chinese | 1382 | 1396 |
| LMArena German | 1380 | 1371 |
| LMArena Korean | 1339 | 1271 |
| LMArena Russian | 1377 | 1331 |
| LMArena Spanish | 1375 | 1372 |
| LMArena French | 1383 | — |
| LMArena Japanese | 1320 | — |
Instruction Following Too close to call
GPT-4.1: 71.5 (#148), Step 3: 70.6 (#159)
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| LMArena Instruction Following | 1367 | 1332 |
| IFEval | 83.8% | — |
Long Context Too close to call
GPT-4.1: 40.0 (#158), Step 3: 40.3 (#152)
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| LMArena Longer Query | 1385 | 1326 |
| Fiction.LiveBench | 63.9% | — |
Writing & Preference GPT-4.1 leads
GPT-4.1: 57.8 (#120), Step 3: 54.5 (#146)
| Benchmark | GPT-4.1 | Step 3 |
|---|---|---|
| LMArena Text | 1383 | 1351 |
| LMArena Creative Writing | 1363 | 1322 |
| LMArena Multi-Turn | 1398 | 1340 |
| EQ-Bench Creative Writing | 1420 | — |
| WildBench | 85.4% | — |
Frequently asked questions
Is GPT-4.1 better than Step 3?
Step 3 is the stronger model overall, scoring 39.9 to 35.7 on the Noometry Index.
Is GPT-4.1 or Step 3 better for coding?
Step 3 scores higher on coding benchmarks: 39.5 versus 34.2 in the Noometry coding category.
How many benchmarks do GPT-4.1 and Step 3 share?
17 benchmarks have published results for both models. GPT-4.1 has 52 scored results on Noometry and Step 3 has 17.