LM Studio has incorporated Z.ai’s newly released GLM-5.3-Flash model into Bionic, its AI agent platform, introducing multimodal input capabilities, a 1 million-token context window, and substantially reduced pricing compared to GLM-5.2. Here are the key details.
LM Studio continues rapidly expanding the models and functionalities available in Bionic, its macOS and Windows platform designed for agentic tasks such as programming, research, and document/file management.
Bionic supports both locally executed models and cloud-hosted versions running on US-based servers, with the latter governed by a strict zero-data-retention policy.
Shortly after the launch of LM Studio Bionic, the company introduced support for Moonshot AI’s Kimi K3. Now, the firm has enabled cloud access to Z.ai’s new GLM-5.3-Flash model, which LM Studio claims is up to ten times less expensive to operate than GLM-5.2.
LM Studio announced this development just hours after the official debut of GLM-5.3-Flash, which had already garnered significant attention through anonymous evaluations on OpenCode and OpenRouter under the internal codename “Ox Alpha.”
GLM-5.3-Flash is a 320-billion parameter Mixture-of-Experts architecture featuring 18 billion active parameters, supporting both image and text inputs, and offering a one-million-token context window.
It outperforms GLM-5.2 across the benchmarks presented by Z.ai, while typically performing similarly to leading frontier models from Anthropic, OpenAI, Google, and DeepSeek.
For further guidance on utilizing GLM-5.3-Flash within LM Studio Bionic, visit this link.
Do you maintain local models on your Mac? Please share your experiences in the comments section.
Recommended reading on Amazon