Model comparison
GPT-6.1 Sol vs Step 3.5 Flash
GPT-6.1 Sol is the stronger model overall, scoring 65.6 to 42.3 on the Noometry Index. Step 3.5 Flash costs 27× less per token, which makes it the better buy when GPT-6.1 Sol's lead doesn't matter for your workload.
Last verified . 13 shared benchmarks.
StepFunSummary
- They share 13 benchmarks with published results for both. GPT-6.1 Sol scores higher in 8 categories and Step 3.5 Flash in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6.1 Sol leads 81.9 to 22.2.
- The biggest single-benchmark swing is NYT Connections (extended): 95.5% for GPT-6.1 Sol and 28.4% for Step 3.5 Flash.
- Step 3.5 Flash is cheaper at $0.10 / $0.30 per million input/output tokens, against $2 / $10 for GPT-6.1 Sol.
- GPT-6.1 Sol accepts more context: 1.05M tokens versus 256K.
- Step 3.5 Flash has downloadable open weights; the other is API-only.
Side by side
| GPT-6.1 Sol | Step 3.5 Flash | |
|---|---|---|
| Provider | OpenAI | StepFun |
| Noometry Index | 65.6 | 42.3 |
| Released | 2026-09-29 | 2026-01-29 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 256K |
| Max output | 128K | 256K |
| Input $ / M tokens | $2 | $0.10 |
| Output $ / M tokens | $10 | $0.30 |
| Results tracked | 34 | 19 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6.1 Sol leads
GPT-6.1 Sol: 63.2 (#8), Step 3.5 Flash: 42.4 (#105)
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| LMArena Coding | 1487 | 1436 |
| DeepSWE | 75.2% | — |
| FrontierCode | 50.2% | — |
| LMArena WebDev | 1755 | — |
| SciCode | 55.8% | — |
Agentic & Tool Use Not comparable
GPT-6.1 Sol: 39.6 (#26), Step 3.5 Flash: —
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| APEX-Agents | 60% | — |
| GDP.pdf | 32% | — |
Reasoning GPT-6.1 Sol leads
GPT-6.1 Sol: 81.9 (#2), Step 3.5 Flash: 22.2 (#202)
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| NYT Connections (extended) | 95.5% | 28.4% |
| LMArena Hard Prompts | 1466 | 1411 |
| ARC-AGI-2 | 94.2% | — |
| ARC-AGI-1 | 98.5% | — |
| CritPt | 31.7% | — |
| Chess Puzzles | 61% | — |
| EBR-Bench | 54.3% | — |
| Mystery Game Puzzles | 80% | — |
| Epoch Capabilities Index | 166.09 | — |
Math GPT-6.1 Sol leads
GPT-6.1 Sol: 93.7 (#1), Step 3.5 Flash: 42.6 (#84)
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| LMArena Math | 1464 | 1408 |
| FrontierMath (Tiers 1-3) | 93.7% | — |
| FrontierMath Tier 4 | 100% | — |
| MathArena Final-Answer Competitions | — | 66.8% |
| OTIS Mock AIME 2024-2025 | 100% | — |
| ProofBench | 99% | — |
Knowledge GPT-6.1 Sol leads
GPT-6.1 Sol: 71.8 (#4), Step 3.5 Flash: 39.6 (#132)
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| LMArena Expert | 1502 | 1421 |
| GPQA Diamond | 95.4% | — |
| SimpleQA Verified | 73.9% | — |
Multimodal Not comparable
GPT-6.1 Sol: 52.7 (#5), Step 3.5 Flash: —
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| LMArena Vision | 1288 | — |
| Furniture Assembly | 80% | — |
Multilingual GPT-6.1 Sol leads
GPT-6.1 Sol: 54.3 (#46), Step 3.5 Flash: 50.5 (#119)
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| LMArena Non-English | 1438 | 1385 |
| LMArena Chinese | 1477 | 1447 |
| LMArena Russian | 1455 | 1385 |
| LMArena French | — | 1421 |
| LMArena German | — | 1405 |
| LMArena Japanese | — | 1354 |
| LMArena Korean | — | 1352 |
| LMArena Spanish | — | 1419 |
Instruction Following GPT-6.1 Sol leads
GPT-6.1 Sol: 77.0 (#29), Step 3.5 Flash: 73.1 (#124)
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| LMArena Instruction Following | 1468 | 1385 |
Long Context GPT-6.1 Sol leads
GPT-6.1 Sol: 44.9 (#54), Step 3.5 Flash: 42.8 (#117)
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| LMArena Longer Query | 1465 | 1402 |
Writing & Preference GPT-6.1 Sol leads
GPT-6.1 Sol: 63.6 (#63), Step 3.5 Flash: 58.8 (#113)
| Benchmark | GPT-6.1 Sol | Step 3.5 Flash |
|---|---|---|
| LMArena Text | 1447 | 1403 |
| LMArena Creative Writing | 1432 | 1357 |
| LMArena Multi-Turn | 1449 | 1405 |
Frequently asked questions
Is GPT-6.1 Sol better than Step 3.5 Flash?
GPT-6.1 Sol is the stronger model overall, scoring 65.6 to 42.3 on the Noometry Index. Step 3.5 Flash costs 27× less per token, which makes it the better buy when GPT-6.1 Sol's lead doesn't matter for your workload.
Which is cheaper, GPT-6.1 Sol or Step 3.5 Flash?
Step 3.5 Flash is cheaper. It lists at $0.10 per million input tokens and $0.30 per million output tokens; GPT-6.1 Sol lists at $2 and $10.
Is GPT-6.1 Sol or Step 3.5 Flash better for coding?
GPT-6.1 Sol scores higher on coding benchmarks: 63.2 versus 42.4 in the Noometry coding category.
Which has the bigger context window?
GPT-6.1 Sol does, with 1.05M tokens against 256K.
How many benchmarks do GPT-6.1 Sol and Step 3.5 Flash share?
13 benchmarks have published results for both models. GPT-6.1 Sol has 34 scored results on Noometry and Step 3.5 Flash has 19.