Qwen3-4B-Thinking
A 4-billion-parameter chain-of-thought model in Alibaba's Qwen3 series. The model belongs to the Qwen3 family and is equipped with chain-of-thought reasoning capabilities for multi-step inference tasks.
Timeline (last 90 days)
A GSM8K experiment checked into the repository uses a job flow of 72 sessions with hermes memory turned off, using Qwen3-32B to play the student persona and Qwen3-4B-Thinking as the policy model being trained.
The reward in that experiment is provided by a PRM (process reward model), which is a frozen base version of the 4B policy model, and the entire process runs on seven GPUs.
All Facts (2)
A GSM8K experiment checked into the repository uses a job flow of 72 sessions with hermes memory turned off, using Qwen3-32B to play the student persona and Qwen3-4B-Thinking as the policy model being trained.
50%UnverifiedThe reward in that experiment is provided by a PRM (process reward model), which is a frozen base version of the 4B policy model, and the entire process runs on seven GPUs.
50%