Step 5 Preview
Output 134 tokens/sFirst token 0.8 sContext 1M tokens
Top 1 cheapest providers (Adjusted $/task)
Within the active global provider, residency and confidentiality filters.
| # | Provider | Price channel | Raw input $/1M | Raw output $/1M | Adjusted $/task |
|---|---|---|---|---|---|
| 1 | Featherlessvia OpenRouter — Price via OpenRouter: OpenRouter's price for Featherless's endpoint (https://openrouter.ai/stepfun/step-5-preview/providers), collected 2026-11-03. | via OpenRouter — Price via OpenRouter: OpenRouter's price for Featherless's endpoint (https://openrouter.ai/stepfun/step-5-preview/providers), collected 2026-11-03. | $1.00 | $2.70 |
Composite
1 of 7 inputs6 radar axes: DesignArena's two boards share one47.2
Thin data · 1/7Based on only 1 of 7 Composite inputs — treat this rank as uncertaindominance-adjusted from 65.3: a better-measured model that is at least as good on each of these inputs ranks above it
AA Coding v4.x —Coding Agent v1.4 —AA Intelligence 28.5AA Agentic —Epoch ECI —Software ECI —DesignArena —/—
Radar: percentile among all models measured on each input; a gap means not measured.
Benchmark sheet
42 of 319 registered benchmark versions · bars show the percentile among all models measured on each benchmark.
40 of 43 values are StepFun's own claims (†), not independent measurements; a matching independent result replaces a claim as soon as one exists.
Agentic
AA-Briefcasedeveloper's claim1,417† self-reported by the developer
AA-Briefcase published 2026-09-20 → · Vendor-reported by StepFun — AA-Briefcase result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Agents' Last Exam (ALE-CLI)developer's claim29.5%† self-reported by the developer
Agents' Last Exam (ALE-CLI) published 2026-09-20 → · Vendor-reported by StepFun — Agents' Last Exam (ALE-CLI) result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Apex-Agentsdeveloper's claim37.8%† self-reported by the developer
Apex-Agents published 2026-09-20 → · Vendor-reported by StepFun — Apex-Agents result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
AutomationBench-AAdeveloper's claim51.0%† self-reported by the developer
AutomationBench-AA published 2026-09-20 → · Vendor-reported by StepFun — AutomationBench-AA result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
AutomationBench (public)developer's claim44.0%† self-reported by the developer
AutomationBench (public) published 2026-09-20 → · Vendor-reported by StepFun — AutomationBench (public) result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
BrowseCompdeveloper's claim88.7%† self-reported by the developer
BrowseComp published 2026-09-20 → · Vendor-reported by StepFun — BrowseComp result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Dracodeveloper's claim83.3%† self-reported by the developer
Draco published 2026-09-20 → · Vendor-reported by StepFun — Draco result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
FinStepBench-LiveSearchdeveloper's claim74.5%† self-reported by the developer
FinStepBench-LiveSearch published 2026-09-20 → · Vendor-reported by StepFun — FinStepBench-LiveSearch result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
HLE w/ toolsdeveloper's claim59.4%† self-reported by the developer
HLE w/ tools published 2026-09-20 → · Vendor-reported by StepFun — HLE w/ tools result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
JobBenchdeveloper's claim59.0%† self-reported by the developer
JobBench published 2026-09-20 → · Vendor-reported by StepFun — JobBench result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
OfficeQA Prodeveloper's claim60.3%† self-reported by the developer
OfficeQA Pro published 2026-09-20 → · Vendor-reported by StepFun — OfficeQA Pro result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
PresentBenchdeveloper's claim76.8%† self-reported by the developer
PresentBench published 2026-09-20 → · Vendor-reported by StepFun — PresentBench result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
SpeadSheet v2developer's claim29.4%† self-reported by the developer
SpeadSheet v2 v2 → · Vendor-reported by StepFun — SpeadSheet v2 result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
τ³-Bankingdeveloper's claim42.5%† self-reported by the developer
τ³-Banking published 2026-09-20 → · Vendor-reported by StepFun — τ³-Banking result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Terminal-Bench v2.1developer's claim85.0%† self-reported by the developer
Terminal-Bench v2.1 v2.1 → · Vendor-reported by StepFun — Terminal-Bench v2.1 result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Terminal-Bench v4 v4.0developer's claim33.3%† self-reported by the developer
Terminal-Bench v4 v4.0 → · Vendor-reported by StepFun — Terminal-Bench v4 result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Coding
Artificial Analysis Coding Agent Index v1.53351.2%
Artificial Analysis Coding Agent Index v1.5 v1.5 → · Claude Code — Measures coding-agent systems on DeepSWE v1.1, Terminal-Bench 4.0 and SWE-Atlas-QnA.
Artificial Analysis Coding Agent Index v2.04043.1%
Artificial Analysis Coding Agent Index v2.0 v2.0 → · Claude Code — Measures coding-agent systems on FrontierCode v1.1 Main, Terminal-Bench 4.0 and SWE-Atlas-QnA-AA.
DeepSWE v1.1developer's claim67.7%† self-reported by the developer
DeepSWE v1.1 v1.1 → · Vendor-reported by StepFun — DeepSWE v1.1 result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
MLS-Bench-Litedeveloper's claim40.5%† self-reported by the developer
MLS-Bench-Lite published 2026-09-20 → · Vendor-reported by StepFun — MLS-Bench-Lite result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
ProgramBench (Pass Rate)developer's claim80.5%† self-reported by the developer
ProgramBench (Pass Rate) published 2026-09-20 → · Vendor-reported by StepFun — ProgramBench (Pass Rate) result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
RoadmapBenchdeveloper's claim54.3%† self-reported by the developer
RoadmapBench published 2026-09-20 → · Vendor-reported by StepFun — RoadmapBench result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
SciCodedeveloper's claim58.9%† self-reported by the developer
SciCode published 2026-09-20 → · Vendor-reported by StepFun — SciCode result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
StepCode-Bench-Dailydeveloper's claim64.9%† self-reported by the developer
StepCode-Bench-Daily published 2026-09-20 → · Vendor-reported by StepFun — StepCode-Bench-Daily result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
StepCode-Bench-Generaldeveloper's claim65.0%† self-reported by the developer
StepCode-Bench-General published 2026-09-20 → · Vendor-reported by StepFun — StepCode-Bench-General result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
StepCodeBenchdeveloper's claim49.0%† self-reported by the developer
StepCodeBench published 2026-09-20 → · Vendor-reported by StepFun — StepCodeBench result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
SWE-Atlas-QnAdeveloper's claim63.6%† self-reported by the developer
SWE-Atlas-QnA published 2026-09-20 → · Vendor-reported by StepFun — SWE-Atlas-QnA result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
SWE-Atlas-Test-writingdeveloper's claim50.8%† self-reported by the developer
SWE-Atlas-Test-writing published 2026-09-20 → · Vendor-reported by StepFun — SWE-Atlas-Test-writing result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
SWE-Marathon v1.1 (Partial Score)developer's claim72.7%† self-reported by the developer
SWE-Marathon v1.1 (Partial Score) v1.1 → · Vendor-reported by StepFun — SWE-Marathon v1.1 (Partial Score) result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Knowledge
FinStepBench-CorporateValuationdeveloper's claim60.6%† self-reported by the developer
FinStepBench-CorporateValuation published 2026-09-20 → · Vendor-reported by StepFun — FinStepBench-CorporateValuation result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
FinStepBench-FinanceDRdeveloper's claim55.8%† self-reported by the developer
FinStepBench-FinanceDR published 2026-09-20 → · Vendor-reported by StepFun — FinStepBench-FinanceDR result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
FrontierFinancedeveloper's claim66.4%† self-reported by the developer
FrontierFinance published 2026-09-20 → · Vendor-reported by StepFun — FrontierFinance result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
GDPval-AA v2developer's claim1,571† self-reported by the developer
GDPval-AA v2 v2 → · Vendor-reported by StepFun — GDPval-AA v2 result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Long-context
AA-LCR v1.1developer's claim88.3%† self-reported by the developer
AA-LCR v1.1 v1.1 → · Vendor-reported by StepFun — AA-LCR v1.1 result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Reasoning
CritPtdeveloper's claim20.9%† self-reported by the developer
CritPt published 2026-09-20 → · Vendor-reported by StepFun — CritPt result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
GPQA Diamonddeveloper's claim93.5%† self-reported by the developer
GPQA Diamond published 2026-09-20 → · Vendor-reported by StepFun — GPQA Diamond result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
HLEdeveloper's claim46.5%† self-reported by the developer
HLE published 2026-09-20 → · Vendor-reported by StepFun — HLE result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
AA Intelligence Index v5.0 · snapshot-2026-11-034428.5
AA Intelligence Index v5.0 · snapshot-2026-11-03 → · Published board — Artificial Analysis Intelligence Index v5.0. Models AA has not rescored on v5.0 are listed as estimates from their frozen v4.x value (isotonic fit over the models measured on both versions), never as measurements.
Safety/Alignment
CyberGymdeveloper's claim84.7%† self-reported by the developer
CyberGym published 2026-09-20 → · Vendor-reported by StepFun — CyberGym result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Tool-use
MCP-Atlasdeveloper's claim85.6%† self-reported by the developer
MCP-Atlas published 2026-09-20 → · Vendor-reported by StepFun — MCP-Atlas result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Toolathlon-Verifieddeveloper's claim74.1%† self-reported by the developer
Toolathlon-Verified published 2026-09-20 → · Vendor-reported by StepFun — Toolathlon-Verified result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Vision
GDP.pdfdeveloper's claim14.8%† self-reported by the developer
GDP.pdf published 2026-09-20 → · Vendor-reported by StepFun — GDP.pdf result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
MMMU-Prodeveloper's claim76.0%† self-reported by the developer
MMMU-Pro published 2026-09-20 → · Vendor-reported by StepFun — MMMU-Pro result as reported by StepFun in the Step-5 Preview launch table; this registry identity preserves the printed version or a dated snapshot when none was stated.
Missing coverage · 277 benchmark versions
No result does not mean a zero, or that the model was never tested. Collection failures and disputed versions retain their distinct status.
Token offers by platform · 1 offer (Adjusted $/task)
Click any underlined price to see how it is estimated and where each input comes from. How we calculate adjusted cost.
1 offer within the active global filters, 2 more outside them; “price n/a” means the provider serves the model but publishes no direct token price we could collect — no other channel's price is shown in its place.
OpenRouter (1)
Prices via OpenRouter — what OpenRouter charges for each provider's endpoint, not the provider's own direct price. Source: openrouter.ai/stepfun/step-5-preview/providers, collected 2026-11-03.
| Featherless | globalfeatherless/fp4 | $1.00 raw in $/1M | $2.70 raw out $/1M |