Skip to main content
Kaino.dev
Discover
Evals
News
Academics
Insights
Kaino.dev

Discover, evaluate, and compare AI tools, models, and agents.

Explore

  • Discover
  • Evaluations
  • News
  • Academics
  • Insights

Community

  • Twitter
  • YouTube
  • Instagram
Privacy PolicyTerms of Service

© 2026 Kaino.dev. All rights reserved.

Version 1.1.0
Step 3.5 Flash · Discover · Kaino
Discover/MODELS/Step 3.5 Flash
Step 3.5 Flash logo

MODELS

Step 3.5 Flash

by StepFun

agenticstepfun
Add to comparisonVisit WebsiteDocumentationGitHub

Overview

Open sparse MoE reasoning model from StepFun with 196B total parameters, 11B active parameters, 256K context, and support for coding, tool use, agentic workflows, deep research, API access, and local deployment.

Details

Step 3.5 Flash is StepFun’s open sparse MoE reasoning model. The official launch page describes it as a 196B-parameter model with 11B active parameters and a 256K context window, built for high-speed reasoning, coding, agentic intelligence, tool use, deep research, and local deployment. StepFun documentation lists it for API usage as model "step-3.5-flash" through the Chat Completion endpoint, and source pages also reference availability through GitHub and Hugging Face.

When to Use

Use when you need an open MoE reasoning model with documented coding tool-use agentic and deep-research capabilities. Use when evaluating models that can be accessed through StepFun’s API while also being available from GitHub and Hugging Face. Use when a long-context model is needed and the officially stated 256K context window fits the task.

Getting Started

  1. Read the StepFun reasoning-model guide for model-specific usage notes.
  2. Use the Chat Completion API endpoint POST https://api.stepfun.ai/v1/chat/completions with model="step-3.5-flash" as shown in StepFun’s API documentation.
  3. Review the StepFun pricing page for current token pricing and rate limits before production use.
  4. Inspect the GitHub or Hugging Face project pages if you plan to evaluate local deployment.

Key Features

  • •Sparse MoE architecture with 196B total parameters and 11B active parameters
  • •according to StepFun’s launch page and technical report.
  • •256K context window as stated on the StepFun launch page.
  • •Documented support for reasoning
  • •coding
  • •agentic workflows
  • •tool use
  • •deep research
  • •and local deployment.
  • •Available through StepFun API documentation
  • •GitHub
  • •and Hugging Face source pages.

Capabilities

  • •reasoning
  • •coding
  • •tool use
  • •agentic workflows
  • •deep research
  • •long-context processing
  • •local deployment
  • •chat completions

Last updated Jul 31, 2026