Step 3.7 Flash
Active Parameters
198B (Estimated)
Context Length
Undisclosed
Modality
Text
Architecture
Mixture of Experts (MoE)
License
Proprietary
Release Date
28 May 2026
Knowledge Cutoff
-
API Pricing (per 1M)
Input: $0.20 · Output: $1.15
Evaluation Benchmarks
Rank
#108
| Benchmark | Score | Rank |
|---|---|---|
Agentic Index | 0.22 | 69 |
Coding Index | 0.40 | 98 |
Intelligence Index | 0.20 | 122 |
Rankings
Overall Rank
#108
Coding Rank
#85
About Step 3.7 Flash
StepFun's high-speed specialized model released May 28, 2026, engineered for rapid real-time multilingual tasks, low-latency agent coordination, and fast conversational inference. Optimized for speed-first deployments where sub-second response times and broad language coverage are paramount, Step 3.7 Flash serves as a lightweight backbone for multilingual pipelines, real-time translation workflows, and latency-sensitive multi-agent coordination tasks.
Technical Specifications
Architecture specifications are undisclosed for proprietary models.
Attention
Attention Structure
Multi-Head Attention
Attention Heads
Undisclosed
Key-Value Heads
Undisclosed
Attention Head Dimension
Undisclosed
Position Embedding
Absolute Position Embedding
RoPE Theta
Undisclosed
Sliding Window Attention
Undisclosed
Sliding Window Size
Undisclosed
Sliding Window Ratio
Undisclosed
Linear Attention
Undisclosed
Linear Attention Ratio
Undisclosed
Normalization
Undisclosed
Activation Function
Undisclosed
Dimensions
Hidden Dimension Size
Undisclosed
Number of Layers
Undisclosed
FFN Intermediate Size (Dense)
Undisclosed
Multi-Token Prediction Heads
Undisclosed
Tokenizer
Vocabulary Size
Undisclosed
Mixture of Experts
Total Expert Parameters
11.0B
Number of Experts
Undisclosed
Active Experts
Undisclosed
Shared Experts
Undisclosed
FFN Intermediate Size (per Expert)
Undisclosed
Dense Layers Before MoE
Undisclosed
Resources
About Step 3.7
StepFun's Step 3.7 generation focuses on high-speed, specialized inference for real-time multilingual tasks, rapid agent coordination, and low-latency conversational applications. The Flash tier is optimized for speed-first deployments where rapid response time and multilingual versatility are the primary requirements.
Other Step 3.7 Models
- No related models available