- Chat
GLM 5.3 Abliterated
A derivative of GLM 5.3 with modified refusal behavior for general-purpose inference.
GLM model family
About this model
GLM 5.3 Abliterated is a modified derivative of GLM 5.3. Abliterated describes a derivative with modified refusal behavior. Responses can still be refused or inaccurate. Arnict's acceptable-use rules apply to every model.
Acceptable use applies equally to modified and original models. The label does not mean unrestricted use or guaranteed responses.
- Parameters
- 753B753.33 billion parameters in the official release tensor metadata.
- Native context
- 1,048,576Published model-card value
- Output guidance
- 163,840Upstream evaluation budget, not a declared maximum or configured API limit.
Capabilities
- Chat completionsUse this model with the chat-completions API when it is marked Available.
- Native input modalitiesText input is listed in the published catalog metadata.
- Long-context input1,048,576 tokens
- Published output guidance163,840 tokens · Upstream evaluation budget, not a declared maximum or configured API limit.
Published limits
Published model-card facts, kept separate from live availability.
| Metric | Published value | Note |
|---|---|---|
| Arnict API context limit | Awaiting connection metadata | Configured serving limit; the native model specification is listed separately |
| Arnict API output limit | Not published | Maximum output accepted by the Arnict API for this model |
| Parameters | 753.33 billion parameters in the official release tensor metadata. | Upstream model card |
| Native model context | 1,048,576 tokens | Upstream model card |
| Output guidance | 163,840 tokens upstream evaluation budget | Upstream evaluation budget, not a declared maximum or configured API limit. |
| Weights / activations | Official release: FP8 E4M3 with dynamic activations. Connected serving format may differ. | Published format |
- Abliterated derivative with modified refusal behavior
- 78 transformer layers
- 256 routed experts; 8 selected per token; 1 shared expert
- 6,144 hidden dimensions
- Text-only input and output
- Reasoning effort: low, high or max
Use zai/glm-5.3-uncensored with an OpenAI-compatible client after this model is enabled.
Related models
GLM
GLM 5.3 Flash Abliterated
A derivative of GLM 5.3 Flash with modified refusal behavior for general-purpose inference.
DeepSeek
DeepSeek V4.1 Flash
Multimodal mixture-of-experts model: 552B backbone parameters, with 8B active during prefill and 16B during decoding.
Qwen
Qwen 3.6 35B A3B
Vision-language mixture-of-experts model: 35B total, 3B active parameters.