GLM 5.3 Flash

Über dieses Modell

ZAI GLM 5.3 Flash is a 320B parameter Mixture-of-Experts (MoE) language model with 18B active parameters, and the first natively multimodal model in the GLM-5 series. It introduces a hybrid architecture combining sparse and linear attention to reduce long-context serving costs, along with Manifold-Constrained Hyper-Connections (mHC) for improved scaling efficiency, trained on a 30T-token multimodal corpus. The model supports configurable reasoning effort levels (low, high, max) with thinking enabled by default, and features a context window of up to 1 million tokens. It outperforms GLM-5.2 across benchmarks and real-world workloads at one-tenth the price, while approaching Claude Opus 4.8 on coding and agentic benchmarks. Released under the MIT license.

Technische Daten

Fähigkeiten

Eingabe-Modalitäten

Ausgabe-Modalitäten

Reasoning

Standard on

Knowledge horizon