Qwen/Qwen3-30B-A3B · Hugging Face
repo
Fields
id
sources/qwen3-30b-a3b-cardKind
repo
Accessed
2026-10-19
notes
Read https://huggingface.co/Qwen/Qwen3-30B-A3B for the benchmark figures on models/qwen3-30b-a3b. As recorded by the reader: The official HF model card for Qwen3-30B-A3B publishes no benchmark numbers (it defers to the blog/report), so scores come from Table 15 (thinking mode) of the official Qwen3 Technical Report: GPQA-Diamond, MATH-500, AIME'24, AIME'25, Arena-Hard, IFEval (strict prompt). The report's coding row is 'LiveCodeBench v5', which is not the allowed 'LiveCodeBench v6', so it was excluded; no MMLU-Pro row exists for the post-trained 30B-A3B.
Notes
Read https://huggingface.co/Qwen/Qwen3-30B-A3B for the benchmark figures on models/qwen3-30b-a3b. As recorded by the reader: The official HF model card for Qwen3-30B-A3B publishes no benchmark numbers (it defers to the blog/report), so scores come from Table 15 (thinking mode) of the official Qwen3 Technical Report: GPQA-Diamond, MATH-500, AIME'24, AIME'25, Arena-Hard, IFEval (strict prompt). The report's coding row is 'LiveCodeBench v5', which is not the allowed 'LiveCodeBench v6', so it was excluded; no MMLU-Pro row exists for the post-trained 30B-A3B.
Cross-references
This record references no other records.