StepFun Releases Step 3.5 Flash: 196B Sparse MoE Model for OpenClaw Agents
By
Zac Zuo
A five-star bake. Worth schmearing, sharing, saving.
Summary
StepFun has released Step 3.5 Flash, a 196B sparse Mixture of Experts (MoE) model that activates only 11B parameters per token for high efficiency. The model delivers frontier reasoning capabilities and strong agentic performance, with seamless native integration for OpenClaw agents. Positioned as one of the best open models for running serious agents, it represents a significant advancement in efficient large language model architecture for agent applications.
Key quotes
· 3 pulledStep 3.5 Flash is StepFun's 196B sparse MoE model that activates only 11B parameters per token.
It delivers frontier reasoning and strong agentic performance with high efficiency.
Seamless native OpenClaw integration makes it one of the best open models for running serious agents right now.
You might also wanna read
GLM-4.7-Flash: Z.ai's 30B-A3B MoE Model for Lightweight AI Deployment
GLM-4.7-Flash is a 30B-A3B Mixture of Experts (MoE) model developed by Z.ai, positioned as the strongest model in the 30B parameter class. T
Step 3.5 Flash Open-Source Foundation Model Performance Benchmarks
The article presents benchmark performance scores for Step 3.5 Flash, described as the most capable open-source foundation model. The conten
static.stepfun.com·3mo ago
Step 3.7 Flash: A High-Efficiency Multimodal AI Model for Real-World Applications
Step 3.7 Flash is a high-efficiency AI model designed for real-world applications. It features native multimodal understanding and acting ca
static.stepfun.com·2d ago
Pangu Pro MoE: Mixture of Grouped Experts for Efficient Sparsity
arxiv.org·11mo ago
Liqvid AI Releases LFM2.5-8B-A1B: An Edge MoE Model for Tool Calling on Consumer Hardware
Liqvid AI released LFM2.5-8B-A1B, an edge-optimized Mixture-of-Experts model designed for fast, reliable tool calling on consumer hardware.
liquid.ai·2d ago
DeepSeek-V4 Series Preview: Million-Token Context MoE Models with 1.6T Parameters
DeepSeek introduces the V4 series of Mixture-of-Experts (MoE) language models, including DeepSeek-V4-Pro (1.6T parameters, 49B activated) an