VerdictPal
Join

Models / StepFun

Step 3.7 Flash

Step · Released 2026-06-01

StepFun's high-efficiency multimodal Flash model, evaluated by Artificial Analysis on 1 Jun. Public catalogs describe image/video input and a 256K context window.

Efficiency-first multimodal model for routing, visual extraction, and cheap agent steps before escalating to a larger model.

#35 of 51 on AA Index · current snapshot
30 AA Index · 10 benchmark rows · 1 sourcesPosition is benchmark-specific — not a cross-family or cross-source ranking.
Composite44
Context256K
Input / 1M—
Output / 1M—
Knowledge cutoffopen
Output speed401 t/s
TTFT (default API)0.72s
Statuslightweight
ReasoningCodingAgenticLong context
Reasoning
81
Coding
40
Agentic
39
Long context
64

56average across 4 tested families

Benchmark placements

Where Step 3.7 Flash places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.

BenchmarkFamilyRankScoreSourceDate
AA output speedPerformance#1/ 69Artificial Analysis2026-07-21
IFBenchReasoning#36/ 5767Artificial Analysis2026-07-21
AA-LCRLong context#42/ 6464Artificial Analysis2026-07-21
Humanity's Last ExamReasoning#46/ 6520Artificial Analysis2026-07-21
Artificial Analysis Intelligence IndexReasoning#49/ 6730Artificial Analysis2026-07-21
GPQA DiamondReasoning#50/ 6981Artificial Analysis2026-07-21
SciCodeCoding#51/ 6540Artificial Analysis2026-07-21
AA time to first tokenPerformance#52/ 69Artificial Analysis2026-07-21
Terminal-BenchAgentic#53/ 6939Artificial Analysis2026-07-21
τ³-Bench BankingAgentic#55/ 6411Artificial Analysis2026-07-21
Family context

The three highest-scoring carded models in each capability family. Where Step 3.7 Flash shows up, it's highlighted.

ReasoningStep 3.7 Flash · 81
  1. 1
    Claude Mythos Preview
    95
  2. 2
    GPT-5.6 Sol
    94
  3. 3
    Kimi K3
    94
CodingStep 3.7 Flash · 40
  1. 1
    GPT 5.5
    96
  2. 2
    Claude Opus 4.8
    96
  3. 3
    Claude Opus 4.6
    96
AgenticStep 3.7 Flash · 39
  1. 1
    GPT 5.5
    100
  2. 2
    Claude Opus 4.8
    100
  3. 3
    Claude Fable 5
    100
Long contextStep 3.7 Flash · 64
  1. 1
    Kimi K3
    75
  2. 2
    GPT-5.1
    75
  3. 3
    GPT-5.6 Sol
    74
Newest receipts

Benchmark rows added to the public ledger for Step 3.7 Flash in the last 120 days. Older rows live in the full table below.

10 recent
BenchmarkFamilyScoreSourceDays ago
AA-LCRLong context63.67%Artificial Analysis2
AA output speedPerformance401 t/sArtificial Analysis2
AA time to first tokenPerformance0.72sArtificial Analysis2
Artificial Analysis Intelligence IndexReasoning30.3%Artificial Analysis2
GPQA DiamondReasoning80.9%Artificial Analysis2
Humanity's Last ExamReasoning19.9%Artificial Analysis2
IFBenchReasoning67.28%Artificial Analysis2
SciCodeCoding40%Artificial Analysis2
τ³-Bench BankingAgentic11.34%Artificial Analysis2
Terminal-BenchAgentic39.33%Artificial Analysis2
Family coverage

A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.

Reasoning81
Coding40
Agentic39
Long context64
Price vs. performance
0255075100$0.10$0.30$1$3$10$30input USD / 1M tokens · log scale →← intelligence

Hover or tab any dot for name, score, and input price.

AnthropicOpenAIMistralAlibabaDeepSeekGoogleMoonshotMiniMaxxAIMetaZhipuXiaomiMeituanCohere

Full benchmark ledger

Every catalog benchmark for Step 3.7 Flash. Scores link to the original source; gaps mean no public row exists yet.

10 sourced rows

10 of 32 catalog benchmarks have a sourced row for Step 3.7 Flash.

Reasoning

Coding

Agentic

Long context

Performance

Sources