GLM-5.3-Flash (Ox Alpha)
#20 in Multimodal ModelsZhipu Ai · v5.3 · flash (ox alpha) · 2× · last seen Aug 29, 2026
31
Momentum
GLM-5.3-Flash is an Open Source multimodal Sparse-MoE model developed by Zhipu AI (Z.ai) with 320 billion total parameters and 18 billion active parameters per Token. It was initially tested anonymously under the codename "Ox Alpha" on OpenRouter and OpenCode before Zhipu officially introduced it as GLM-5.3-Flash on August 26, 2026. The model is the first natively multimodal model in the GLM-5 series, supports text, image, and video inputs, features a context window of approximately 1 million Token, and is distributed as an MIT-licensed open-weight model on Hugging Face.
Momentum trend
01.06.30.08.
Features
| Key Benchmark (%) | 63.4% Pass@1 on DeepSWE v1.1 (Z.ai official figure); 84.3 on Terminal-Bench 2.1 |
| Context Window (Tokens) | 1,048,576 tokens (~1M), max output 131,072 tokens |
| License | MIT License (open model weights) |
| Multimodality | Natively multimodal: text, image, video, visual documents and interleaved multimodal inputs |
| Platform | Z.ai General API, GLM Coding Plan, Hugging Face (open weights), also available via OpenRouter and other third-party providers |
| Price per 1M Tokens | List price: $0.15 input / $0.50 output per 1M tokens (currently 50% launch discount until Sept 9, 2026: $0.075 input / $0.25 output) |
| Release Date | August 26, 2026 (official reveal; previously previewed anonymously as 'Ox Alpha' from August 20, 2026) |