StepFunReasoning modelStep 5 Preview
Step 5 Preview is a StepFun proprietary reasoning model, released 18 Sept 2026. Its best configuration ranks #92 of 756 on the BenchLeader Index at 61.6 ±5.0, in the upper half. The ± is the point: 186 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in composite (80) and lowest in agents & tools (48). At $1.43 per million tokens blended it is pricier than most ranked models. Output speed of 86 tokens per second puts it faster than most, with a first answer in 26.4 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 2 Oct 2026.
- Blended price
- $1.43/M
- $1.00 in · $2.70 out
- Output speed
- 86 tok/s
- measured by Artificial Analysis
- First answer
- 26 s
- Context
- 1M
- Full answer
- 32 s
- median, reasoning included
- Cost per run
- $0.717
- one full Intelligence Index run
- Released
- 18 Sept 2026
- Overall index62
- Reasoning73
- Coding66
- Agents & tools48
- Maths51
- Knowledge69
- Human preference64
- Multimodal61
- Long context70
- Composite80
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| high | – | – | 86 tok/s | 26 s | $0.0012 | Coding 61 |
| not statedbest | 61.6 | #92 | 86 tok/s | 26 s | $0.0012 | Agents & tools 48 · Coding 66 · Composite 80 · Human preference 64 · Knowledge 69 · Long context 70 · Maths 51 · Multimodal 61 · Reasoning 73 |
How its index has moved
19 Sept 2026 to 5 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LMArena Hard Prompts | – | 1482#62 | LMArena | |
| Humanity's Last Exam (AA)not in index | – | 46.5%#40 | Artificial Analysis | |
| CritPt | – | 20.9%#51 | Artificial Analysis |
Coding
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| SciCode | – | 58.9%#15 | SciCode | |
| LMArena Coding | – | 1511#58 | LMArena | |
| LMArena WebDev | 1570#32 | – | LMArena | |
| SciCode (AA)not in index | – | 58.9%#15 | Artificial Analysis | |
| Terminal-Bench 4.0 (AA)not in index | – | 33.3%#35 | Artificial Analysis |
Agents & tools
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LMArena Agent | – | 0.5#29 | LMArena | |
| GDPval (AA)not in index | – | 53.3%#33 | Artificial Analysis | |
| ITBench SRE (AA)not in index | – | 55.6%#2 | Artificial Analysis | |
| Analyst Agent (AA)not in index | – | 35.0%#19 | Artificial Analysis | |
| APEX-Agents (AA)not in index | – | 38.0%#5 | Artificial Analysis | |
| AutomationBenchnot in index | – | 51.0%#37 | Artificial Analysis | |
| GDP.pdfnot in index | – | 14.8%#36 | Artificial Analysis | |
| Harvey LABnot in index | – | 93.4%#6 | Harvey | |
| EnterpriseOps-Gymnot in index | – | 47.2%#8 | Artificial Analysis | |
| AA-Briefcasenot in index | – | 1424#33 | Artificial Analysis |
Maths
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| ProofBench | – | 42.0%#35 | Vals AI |
Knowledge
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| AA-Omniscience | – | 16.4#75 | Artificial Analysis | |
| AA-Omniscience: accuracynot in index | – | 41.5%#106 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | – | 57.0%#78 | Artificial Analysis |
Instruction following
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LMArena Instruction Followingnot in index | – | 1445#71 | LMArena |
Human preference
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LMArena Text | – | 1454#72 | LMArena | |
| LMArena Creative Writingnot in index | – | 1421#82 | LMArena | |
| LMArena Multi-turnnot in index | – | 1460#76 | LMArena | |
| LMArena Longer Queriesnot in index | – | 1468#62 | LMArena |
Multimodal
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LMArena Vision | – | 1278#36 | LMArena | |
| MMMU-Pro | – | 76.4%#84 | Artificial Analysis |
Long context
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| AA-LCR | – | 88.3%#2 | Artificial Analysis | |
| MLCRnot in index | – | 16.7%#24 | Artificial Analysis |
Composite
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| AA Intelligence Index | – | 43.7#41 | Artificial Analysis |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.048 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0012 | $0.0009 | 29.9 s |
| Summarise a 30-page report | 12,000 / 600 | $0.014 | $0.0051 | 33.3 s |
| Code edit | 6,000 / 1,500 | $0.010 | $0.0058 | 43.8 s |
| Agentic coding session | 60,000 / 4,000 | $0.071 | $0.028 | 1.2 min |
| Structured extraction | 2,000 / 200 | $0.0025 | $0.0011 | 28.7 s |
See also
Data as of 5 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Step 5 Preview: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/step-5, data as of 5 Oct 2026.