GLM-4.5V

OpenRouter's product page for Z.ai's GLM 5.3 describes it as a large-scale reasoning model built for complex software engineering and long-horizon agent tasks, supporting text input and output with a 1M-token context window. The page notes improvements over GLM-5.2 in coding and in the balance between performance and t OpenRouter routes GLM 5.3 requests to multiple providers including Reka AI, AkashML, Wafer, DeepInfra, Morph, Friendli, Makora, Cloudflare, Fireworks, Baseten (fp4 and fp8), GMICloud, Modal, Parasail, NovitaAI, DigitalOcean, Together, Phala, SiliconFlow, Decart, Inceptron, Sail Research, Z.ai, and io.net. Performance v

Z.ai's official developer release notes document a 2026 release cadence led by GLM-5.x models. GLM-5.3-Flash (2026-08-26) is described as a native multimodal model with visual capabilities spanning interfaces, rendering results, and interaction feedback, using an efficient hybrid architecture combining linear and spars Earlier entries in the timeline include GLM-5.2 (2026-06-16) with 1M lossless context for long-horizon tasks, and GLM-5.1 (2026-04-07) designed for autonomous runs of up to 8 hours with planning, execution, iterative refinement, and delivery capabilities. The release log also notes GLM-5.3's emergent cybersecurity capa

OpenRouter lists Z.ai's GLM-5V-Turbo as a native multimodal agent foundation model built for vision-based coding and agent-driven tasks, natively handling image, video, and text inputs. The page describes the model as excelling at long-horizon planning, complex coding, and task execution, enabling a "perceive → plan → The model is hosted on OpenRouter by a single provider, Z.ai, with P50 latency of 5.68 seconds and throughput of 28 tokens per second, plus a 79.07% average cache hit rate and 4.15% tool call error rate. OpenRouter's pricing data shows a weighted average effective input price of $0.5331 per 1M tokens against the $1.20

Z.ai's GLM-4.5V is hosted on SiliconFlow with metadata listing it as part of the GLM-V family, based on ZhipuAI's GLM-4.5-Air foundation model. According to SiliconFlow's model page, GLM-4.5V achieves state-of-the-art performance on image, video, and document understanding as well as GUI agent operations, and is design SiliconFlow's listing specifies that GLM-4.5V uses a Mixture-of-Experts Transformer architecture with 106B total parameters and 12B activated parameters, FP8 precision, calibrated status, a 66K context length, and no reasoning mode. Critically, SiliconFlow marks the model's state as "Deprecated," indicating it is no lo