Unverified50% confidenceFactExact time
Step 5 Preview was specially trained for coding, financial research and long-running agent scenarios.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/17/2026
First Seen
Valid until: 1/15/2027
Sources
Step 5 Preview: A 600B Open-Source MoE Model Built for Long-Running Agents
producthuntProductHunt Daily10/17/2026
Related Entities
Related Claims
UnverifiedStep 5 Preview's product page does not disclose the composition of its training data, the specific evaluation methods for long-running agents, or comparison data with similar models on long-horizon tasks.78% similarUnverifiedThe pretraining and fine-tuning paradigm originated from NLP and computer vision, with its core being large-scale data pretraining to learn general low-level features, and downstream tasks adapted using a small amount of labeled data.71% similarUnverifiedGLM-5.2 团队大幅扩展了面向编码智能体场景的 1M 上下文训练，覆盖大规模实现、自动化研究、性能优化和复杂调试等任务类型70% similarUnverifiedThe training pipeline involves first training an image baseline model, using the image model to warm-start an eight-frame video baseline model, and ultimately generating 5.6-second inference outputs.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1053348API
curl https://kongchang.com/api/v1/knowledge/claims/1053348MCP
get_claim(id=1053348)