AIEvals

Models

Step 5 Preview

Specifications and API prices for Step 5 Preview, recorded by Artificial Analysis.

StepFunReleased 18 September 2026Context 1M tokensOutput Not published in the checked sourcePrivate weights$1.00 in / $2.70 out per Mtok

Family Step14 benchmarks with a published resultNo first place on the independent columns among the 46 models this site listsRead 17 October 2026

Where it stands

Rank compares tracked models with a score. Published shows the source’s rank and field size, when available.

BenchmarkScoreRankPublishedLeaderBehind
Artificial Analysis Intelligence IndexArtificial Analysis4416of 42no list readClaude Opus 5.514 pts
Terminal-Bench 4.0Artificial Analysis33.33%15of 40no list readClaude Sonnet 5.530.31 pts
Terminal-Bench Science 0.1Artificial Analysis2.38%22of 24no list readGPT-6 Astra60.95 pts
MLCR-AAArtificial Analysis16.67%27of 37no list readClaude Sonnet 5.558.33 pts
AutomationBench-AAArtificial Analysis50.96%29of 40no list readGemini 4 Argon26.55 pts
SciCodeArtificial Analysis58.91%9of 40no list readClaude Opus 5.57.99 pts
GDP.pdfArtificial Analysis14.80%32of 40no list readGPT-6.1 Sol16.20 pts
CritPtArtificial Analysis20.86%17of 42no list readGPT-5.6 Sol11.43 pts
AA-OmniscienceArtificial Analysis16.3824of 42no list readClaude Opus 5.530.04
AA-LCR v1.1Artificial Analysis88.33%2of 42no list readKimi K30.34 pts
GDPval-AA v2.1Artificial Analysis1,566.1222of 42no list readClaude Opus 5.5280.05
AA-Briefcase v1.1Artificial Analysis1,432.0223of 41no list readClaude Opus 5.5389.83
Humanity's Last ExamArtificial Analysis46.48%18of 42no list readClaude Opus 5.514.87 pts

Nearby on Artificial Analysis Intelligence Index

The two models above and below on the Artificial Analysis Intelligence Index (Artificial Analysis), on every column where this model has a published number.

ModelArtificial Analysis Intelligence IndexindexTerminal-Bench 4.0%Terminal-Bench Science 0.1%
Grok 4.6xAI4417.17%6.19%
Kimi K3Moonshot AI4412.63%–
Step 5 PreviewStepFun4433.33%2.38%
Claude Haiku 5.5Anthropic4332.83%20.00%
Claude Opus 4.8Anthropic4221.72%–
Bold marks a figure better than this model’s on the same column, whichever model it belongs to. Ranks cover the models shown from the stated source.

Cost and speed, as published

Additional measurements from Artificial Analysis’ own row. Read 30 September 2026.

As publishedCost per taskOutput speed
Step 5 Preview$0.7285.4 tok/s

Results by benchmark

Independent and vendor results retain their own harnesses and effort settings.

BenchmarkIndependent runAs reported by a publisher
MLCR-AAMedical reasoning16.67%Artificial Analysisnot published
AutomationBench-AAWorkflow automation50.96%Artificial Analysisnot published
SciCodeScientific coding58.91%Artificial Analysisnot published
GDP.pdfDocument reasoning14.80%Artificial Analysisnot published
CritPtScientific reasoning20.86%Artificial Analysisnot published
AA-OmniscienceFactual knowledge16.38Artificial Analysisnot published
AA-LCR v1.1Long-context reasoning88.33%Artificial Analysisnot published
Terminal-Bench 4.0Agentic coding33.33%Artificial Analysisnot published
MMMU ProMultimodal reasoning
76.36%Artificial Analysis
not published
Terminal-Bench Science 0.1Agentic science2.38%Artificial Analysisnot published
GDPval-AA v2.1Knowledge work1,566.12Artificial Analysisnot published
AA-Briefcase v1.1Knowledge work1,432.02Artificial Analysisnot published
Humanity's Last ExamMultidisciplinary reasoning46.48%Artificial Analysisnot published
Artificial Analysis Intelligence IndexComposite44Artificial Analysisnot published
Independent column: the suite named under each figure, one harness per suite. Publisher column: each lab’s own table, reproduced on the benchmark’s page. Publisher ranks compare reported scores in the same table row only; ties share a rank and unreported cells are excluded.

Sources

Read on 17 October 2026.