Step 3.7 Flash
Step 3.7 Flash is a sparse Mixture-of-Experts (MoE) vision-language model released by StepFun, featuring multimodal understanding capabilities for processing combined image and text inputs. The model is designed with inference efficiency as a primary goal, targeting intelligent agent workloads that require high concurrency and low latency, while integrating language understanding, visual perception, and task execution capabilities.
Timeline (last 90 days)
Group C of the AI World Cup Legacy Bug Challenge features four models: Kimi K3, Grok 4.6, DeepSeek V4 Pro, and Step 3.7 Flash
In the Silver BUG (S003) problem, all four contestants—Kimi K3, Grok 4.6, DeepSeek V4 Pro, and Step 3.7 Flash—answered correctly
In the Gold BUG, both Kimi K3 and Step 3.7 Flash exhausted their two rounds of chances without fixing the bug, scoring nothing
In the first round of verification for the Diamond BUG, all four contestants failed
Grok 4.6 and DeepSeek V4 Pro each accumulated 2 points, while Kimi K3 and Step 3.7 Flash accumulated 0 points
Providing answers too quickly often indicates a misjudgment of the problem's complexity, and reasonable planning (To-Do List, sub-agents) is a prerequisite for success in complex tasks