GLM 5.3 Flash Abliterated
Z AI's GLM-5.3-Flash with refusal behavior removed from the weights. 1M-token context, $0.12 in and $0.50 out per million tokens.
Specifications
- Model id
glm-5.3-flash-abliterated- Base model
- Z AI GLM-5.3-Flash, released 2026-08-26
- Context
- 1,048,576 tokens
- Max output
- 131,072 tokens
- Input
- $0.12 per 1M tokens
- Cached input
- $0.025 per 1M tokens
- Output
- $0.50 per 1M tokens
- Speed
- About 120 tokens per second
- Weights
- NVFP4
- Where it runs
- Refuseless servers only. Prompts and outputs are not stored.
Best for: Everyday and high-volume work. Close to GLM 5.3 on coding benchmarks, for less per token.
Abliteration finds the direction inside the model that produces refusals and removes it from the weights. The model keeps what the base model knows, and nothing is added to your prompt, so a new session does not bring the refusals back. It can still be wrong, and it does not decide what your use is allowed to be: the terms do. How abliteration works
Scores for the base model
DeepSWE: 63.4 (agentic coding). Terminal Bench 2.1: 84.3 (terminal tasks). AutomationBench: 48.8 (business workflows).
These are Z AI's published numbers for GLM-5.3-Flash, from Z AI's GLM-5.3-Flash release post. We have not rerun them on the abliterated weights.
Call it
Send an OpenAI chat completions request to https://api.refuseless.com/v1 with model set to glm-5.3-flash-abliterated. The Anthropic SDK works too, through /v1/messages. SDK setup
curl
curl https://api.refuseless.com/v1/chat/completions \
-H "Authorization: Bearer $REFUSELESS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "glm-5.3-flash-abliterated",
"messages": [{"role": "user", "content": "Explain what powershell -nop -w hidden -enc does and why defenders flag it."}]
}'Try GLM 5.3 Flash Abliterated
Create an account, add credit by card or crypto, and call it with your own key. You pay per token.
