Uncensored gpt-oss-120b vs GLM-5.3 Uncensored
Prices per million tokens as of 2026-11-20. Uncensored gpt-oss-120b is the lower-priced build on a 2M-in, 500K-out month.
| Uncensored gpt-oss-120b | GLM-5.3 Uncensored | |
|---|---|---|
| Model id | gpt-oss-120b-uncensored | glm-5.3-uncensored |
| Kind | House build | Uncensored |
| Developer | PentestLLM | Z.ai |
| Input per M | $0.146 | $0.282 |
| Cached input per M | input rate | $0.0799 |
| Output per M | $0.728 | $0.846 |
| Context window | 131,072 tokens | not published |
| Max output | 131,072 tokens | not published |
| Last hour | No traffic in the last hour | Answering |
| Daily check | Steady since Oct 21 | Steady since Nov 17 |
| Worked month | $0.66 | $0.99 |
The month
- Input
- 2,000,000 tokens
- Output
- 500,000 tokens
- Uncensored gpt-oss-120b
- $0.66
- GLM-5.3 Uncensored
- $0.991.5× the other build
Same call, different id
Both answer the OpenAI wire format on one base URL and key. Nothing but the model id changes.
One request on each
curl https://api.pentestllm.ai/v1/chat/completions \ -H "Authorization: Bearer sk-ptl-live-…" \ -H "Content-Type: application/json" \ -d '{"model": "gpt-oss-120b-uncensored", "messages": [{"role": "user", "content": "Hello"}]}'curl https://api.pentestllm.ai/v1/chat/completions \ -H "Authorization: Bearer sk-ptl-live-…" \ -H "Content-Type: application/json" \ -d '{"model": "glm-5.3-uncensored", "messages": [{"role": "user", "content": "Hello"}]}'Questions
Which is cheaper, Uncensored gpt-oss-120b or GLM-5.3 Uncensored?
Uncensored gpt-oss-120b: $0.66 for a month of 2,000,000 input and 500,000 output tokens against $0.99 on GLM-5.3 Uncensored, 1.5× the price, at the catalog rates on 2026-11-20.
What is the difference between an abliterated and an uncensored build?
An abliterated build has the refusal direction removed from the weights; an uncensored build is a fine-tune or a system-level build that answers without declining. Both answer what the stock model refuses; the model pages name each build.
Can I switch between them without changing code?
Yes. Both answer the OpenAI wire format on the same base URL and key; the model id is the only field that changes.
Which one still answers like itself?
The daily real-model check: Uncensored gpt-oss-120b reads "Steady since Oct 21", GLM-5.3 Uncensored reads "Steady since Nov 17" at render.
Every pair is on the comparisons page; every model with all its prices is on the price index.