Microblog

We are excited to introduce MiMo-V2-Pro, Xiaomi's flagship foundation model.

The model features over 1 trillion total parameters and a 1 million context length. MiMo-V2-Pro is deeply optimized for agentic scenarios and is highly adaptable to general agent frameworks like OpenClaw. It ranks among the global top tier in standard benchmarks such as PinchBench and ClawBench.

MiMo-V2-Pro is designed to serve as the brain of agent systems, orchestrating complex workflows, driving production engineering tasks, and delivering results reliably.

Qwen 3.6 Plus Preview is the next-generation evolution of the Qwen Plus series, featuring an advanced hybrid architecture that improves efficiency and scalability. It delivers stronger reasoning and more reliable agentic behavior compared to the 3.5 series. In benchmarks, it performs at or above leading state-of-the-art models.

Designed as a flagship preview, it excels in agentic coding, front-end development, and complex problem-solving.

Note: The model collects prompt and completion data that can be used to improve the model.

MiniMax M2.7 is a next-generation language model designed for autonomous productivity and continuous improvement. By integrating advanced agentic capabilities and multi-agent collaboration, the model can plan, execute, and adjust complex tasks in dynamic environments.

Key benefits:

• Optimized for real-world workflows (live debugging, root cause analysis).
• High benchmark results (56.2% on SWE-Pro, 57.0% on Terminal Bench 2).
• Suitable for full document generation (Word, Excel, PowerPoint).
• Available in VEGA AI for solving complex tasks.

Introducing two new models from OpenAI: GPT-5.4 Nano and GPT-5.4 Mini.

GPT-5.4 Nano is the most lightweight and cost-effective version in the GPT-5.4 family. Optimized for high-speed tasks, classification, and data extraction. Ideal for background processes and real-time systems where minimal latency is critical.

GPT-5.4 Mini offers a balance between performance and cost. It retains the core capabilities of GPT-5.4, excels at coding, tool usage, and multi-step reasoning. Designed for large-scale projects and assistants.

GLM-5 Turbo by Z.ai is a new model designed for fast inference and high performance in agent environments (e.g., OpenClaw scenarios).

The model is deeply optimized for real workflows involving long execution chains. It features improved decomposition of complex instructions, tool usage, and stability during extended tasks. Try GLM-5 Turbo in Auto and Chat modes!

Grok 4.20 Beta is the latest flagship model from xAI with industry-leading speed and agent capabilities.

It combines the lowest hallucination rate on the market with strict prompt adherence, providing consistently accurate and truthful answers. Supports context up to 2 million tokens. You can already test it in our service.

We have added powerful new tools for Auto and Web modes!

• GitHub Search: AI can now directly search repositories, users, and code snippets, making technical problem-solving even faster.
• Polymarket (and Kalshi) Search: AI has learned to retrieve current forecasts, probabilities, and event quotes from prediction markets. Ask for the probabilities of any world events, and VEGA will find the exact data!