Z.ai · Abliterated

GLM 5.3 Uncensored

Abliterated research variant of GLM-5.3: refusal direction removed from the weights (weight orthogonalization, no adapters or runtime code), served unaligned for red-teaming and interpretability research. Same architecture and 1M context as the base model.

glm-5.3-uncensored

List pricing

—

Input: $0.0699 / 1,000,000 tokens
Output: $1.4 / 1,000,000 tokens
Cache read: $0.0689 / 1,000,000 tokens
Cache write: $0.0699 / 1,000,000 tokens

How pricing works

Discount history

No discount history yet.

Model details

Input: text
Output: text
Context window: 1,048,576 tokens
Maximum output: 131,072 tokens
Availability: Unknown
Updated: 11/8/2026, 3:58:06 AM UTC

Capacity is reserved when your request is accepted.

Capabilities

ChatStreamingReasoningTool calling

Supported parameters

frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_ppresence_penaltyreasoningreasoning_effortrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p