StepFunProvisional

Step 3.5 Flash

Step 3.5 / Flash · stepfun-ai/Step-3.5-Flash

Open Chinese agentic model optimized for fast reasoning, coding, and long-context agents.

Objective score

43.5

Confidence

82%

Price evidence

Open weights; hosted pricing varies

Support this model Compare trend Official source

Scenario evidence

Six task-specific rankings

Feature development#14Obj. 43.50 · 0%Debug and repair#4Obj. 61.90 · 0%Refactoring#3Obj. 58.80 · 0%Code review#4Obj. 58.80 · 0%Large-repo agent work#15Obj. 38.70 · 0%Frontend / UI#2Obj. 57.70 · 0%

Evidence state

Signal coverage

Benchmarks: 2
Rank points: 7
Usage: 2
Agent: 0
Sources: 4

Objective evidence

Benchmarks and public signals

Swebench: 74.4
Livecodebench: —
Aider: —
Artificial Analysis: —
Openrouter Usage: 0
Kilo Usage: —

Persistent history

Recent ranking points

Aug 1#33 · 37

Aug 1#33 · 37

Jul 31#33 · 37

Jul 31#33 · 37

Jul 31#33 · 37

Jul 31#33 · 37

Usage signals

OpenRouter0%

Overall0%

Agent and tool signals

Evidence is being collected. Missing data is not treated as zero.

Source trail

Official StepFun model card reports 256K context and coding/agent benchmark claims.Artificial Analysis Coding Agents · ok WebDev Arena · ok Kilo leaderboard · ok Step 3.5 Flash model card · ok