An open-weight model from Alibaba’s Qwen3 family with about 30 billion total parameters and, per its A3B designation, roughly 3 billion active per token (model card). In this corpus it is the subject of Apple’s self-distillation result, where fine-tuning on its own samples with no verifier, teacher or reinforcement learning raised its LiveCodeBench v6 pass@1 from 42.4% to 55.3%, with the gains concentrated on the hardest problems.
AISystem
Qwen3-30B-Instruct
Alibaba's 30-billion-parameter Qwen3 instruct model (reported in the Apple paper as Qwen3-30B-Instruct, the Qwen3-30B-A3B-Instruct-2507 checkpoint), the base model that simple self-distillation lifted from 42.4% to 55.3% on LiveCodeBench.
Sources
Attribution and reporting linked by this record; links do not imply independent verification.
- Welcome to April 5, 2026 (theinnermostloop.substack.com)
Development timeline
Newest reporting dates first. Reporting dates may differ from event dates.
Cite or correct this record
Suggest a correction — opens a GitHub draft for you to review and submit.