SIQ-1-35B model demo targets autoresearch and agent work
SIQ-1-35B is a model made by further training Qwen-35B-A3 with PPO. The training used , meaning the model was rewarded against results that can be checked. In the shared comparison, it beats GLM-5.2 and Qwen-350B on Karpathy’s autoresearch “” task, and its ideas are described as close in feel to Opus4.8.
On “bullshit-bench,” it is said to score above NEX and GPT-5.5. The model and GGUF files are available on , and a demo can be tried on ZeroGPU.
Key points
- SIQ-1-35B is based on Qwen-35B-A3 and trained further with PPO.
- It is presented as beating GLM-5.2 and Qwen-350B on an autoresearch task.
- It is also presented as beating NEX and GPT-5.5 on bullshit-bench.
- The model and GGUF files are available on .
- A ZeroGPU demo lets people try without a .