GLM 5.2 FP8

GLM

Open WeightModel weights are publicly available — can be downloaded and self-hosted

Released February 20261049K context≈ 786,432 words

GLM 5.2 FP8 is a text-in, text-out model running in 8-bit floating point precision, which reduces memory footprint compared to full-precision variants at the cost of some numerical fidelity. It carries the open-weight MIT license, meaning weights are freely available for inspection and deployment. The FP8 quantization makes it practical for resource-constrained environments where fitting a large model into GPU memory is a real concern.

Capabilities

Capability scores are AI-generated based on model documentation, benchmarks, and technical specifications. Learn more

Creative Writing

Moderate

Instruction Following

GLM 5.2 FP8

GLM

Open WeightModel weights are publicly available — can be downloaded and self-hosted

Released February 20261049K context≈ 786,432 words

GLM 5.2 FP8 is a text-in, text-out model running in 8-bit floating point precision, which reduces memory footprint compared to full-precision variants at the cost of some numerical fidelity. It carries the open-weight MIT license, meaning weights are freely available for inspection and deployment. The FP8 quantization makes it practical for resource-constrained environments where fitting a large model into GPU memory is a real concern.

Capabilities

Capability scores are AI-generated based on model documentation, benchmarks, and technical specifications. Learn more

Creative Writing

Moderate

Instruction Following

Glossary

FP8 QuantizationA compression technique that reduces model size by representing weights using 8-bit floating-point numbers instead of higher precision, making it faster and more memory-efficient.FidelityThe degree to which a quantized or compressed model preserves the quality and accuracy of the original full-precision model.Floating Point PrecisionThe number of bits used to represent decimal numbers in a model; lower precision (like 8-bit) uses less memory but may lose some accuracy compared to higher precision (like 32-bit).Full-PrecisionA model using standard 32-bit floating-point numbers to represent weights, providing maximum accuracy but requiring more memory.GPU MemoryThe high-speed memory on a graphics processor used to store and process model weights and computations during inference.MIT LicenseA permissive open-source license that allows free use, modification, and distribution of software with minimal restrictions.Memory FootprintThe amount of RAM or storage space a model requires to run, which is critical for deployment on resource-constrained devices.Numerical FidelityThe accuracy and precision with which a model preserves mathematical calculations; lower fidelity means some precision is lost, often as a trade-off for smaller model size.PrecisionThe level of numerical detail a model uses to represent its internal values; higher precision means more accurate calculations but requires more memory.QuantizationReducing a model's numerical precision (e.g., from 16-bit to 4-bit) to shrink memory usage and speed up inference.Resource-ConstrainedHardware with limited memory, processing power, or battery life, requiring models to be optimized for efficiency.Text-In, Text-OutA model that accepts text as input and produces text as output, without support for images, audio, or other data types.WeightsThe numerical parameters inside a neural network that determine how it processes input and generates output.

Capabilities

Capabilities

Use Case Fit

Glossary