GLM-5.3 Flash (Together AI)

GLM-5.3 Flash is an open-weight release from Z.ai that shipped on August 26, 2026, and quickly spread across a broad set of inference providers, including Together AI alongside OpenRouter, Cloudflare Workers AI, and others. The model accepts text, images, video, and PDF inputs while producing text outputs, and exposes capabilities such as reasoning, tool calling, temperature control, structured output, and attachment handling. Its broad multimodal input surface and open-weight availability make it a practical choice for teams building code assistants, agentic workflows, and document- or video-aware applications that need flexible deployment rather than a single hosted endpoint.

Beyond availability, GLM-5.3 Flash has been positioned as a cost-efficient coding-focused option that competes with frontier-class rivals on benchmarks while undercutting them on price, which helps explain its rapid provider fan-out. Head-to-head measurements of the model serving through Together AI showed a 98.5% task success rate and a 12.45-second end-to-end p99 latency across 600 structured workloads, indicating that real-world agent and extraction pipelines can run reliably on this release. Together AI exposes the model behind an OpenAI-compatible chat-completions interface, so teams that already standardize on that protocol can adopt GLM-5.3 Flash with minimal integration work.