【Model Launch】Kimi-K3.1-Pro
- Model ID:
moonshotai/kimi-k3.1-pro - Model description: Kimi K3.1 Pro is the first model in the K3.1 series and Kimi’s most capable 3T-scale open-weight model to date. It supports a context window of 1 million tokens and delivers significant improvements in coding, 3D creation, financial research, and a range of agentic tasks. Stronger long-context understanding, complex reasoning, and task execution make it well suited to software development, specialized research, and long-horizon agent applications.
【Model Launch】DeepSeek-V4.1-Flash-1017
- Model ID:
deepseek-ai/deepseek-v4.1-flash-1017 - Model description: DeepSeek-V4.1-Flash-1017 is a minor version update to DeepSeek-V4.1-Flash, with significant improvements in reasoning, coding, agentic tasks, and vision benchmarks. It further reduces task execution time and output length, helping you complete tasks faster and at lower cost. Across multiple agent frameworks, it also expands the Pareto frontier between model performance and usage cost.
【Model Launch】IQuest-Q1
- Model ID:
iquestLab/iquest-q1 - Model description: IQuest-Q1 is designed for agentic coding, complex reasoning, and multi-step tool use. It supports a 512K context window and is suitable for code development, terminal-based agents, long-horizon complex task execution, and long-context analysis.
【Model Launch】DeepSeek-V4.1-Flash
- Model ID:
deepseek-ai/deepseek-v4.1-flash - Model description: DeepSeek-V4.1-Flash is designed for coding assistance, terminal and computer-use agents, long-horizon complex tasks, and long-context analysis. It also natively supports image understanding, making it suitable for applications that need to balance response speed, task performance, and cost efficiency.
【Model Launch】DeepSeek-V4-Flash-0731
- Model ID:
deepseek-ai/deepseek-v4-flash-0731 - Model description: DeepSeek-V4-Flash-0731 is designed for general chat, coding assistance, and complex task processing scenarios. It is suitable for applications that need to balance response speed and reasoning quality.
【Model Launch】GLM-5.2 PD Disaggregated Version
- Model ID:
glm-5.2 - Model description: The GLM-5.2 PD disaggregated version uses a Prefill / Decode disaggregated deployment architecture. It is suitable for long-context, high-throughput, or stability-sensitive inference workloads.