inference/research
verifiedmedium · 0.80updated 2026-10-19

Qwen/Qwen3-Next-80B-A3B-Instruct · Hugging Face

repo

Fields

id
sources/qwen3-next-80b-a3b-instruct-card
Kind
repo
Accessed
2026-10-19
notes
Read https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Instruct for the benchmark figures on models/qwen3-next-80b-a3b-instruct. As recorded by the reader: Read from the 'Performance' table on the official HF model card (instruct/non-thinking). Caveats: the card labels the math row 'AIME25' (reported as AIME 2025) and the alignment row 'Arena-Hard v2' (reported as Arena-Hard); the knowledge row is labelled only 'GPQA' (72.9), not 'GPQA Diamond', so it was excluded rather than substituted. LiveCodeBench row is 'v6 (25.02-25.05)'.

Notes

Read https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Instruct for the benchmark figures on models/qwen3-next-80b-a3b-instruct. As recorded by the reader: Read from the 'Performance' table on the official HF model card (instruct/non-thinking). Caveats: the card labels the math row 'AIME25' (reported as AIME 2025) and the alignment row 'Arena-Hard v2' (reported as Arena-Hard); the knowledge row is labelled only 'GPQA' (72.9), not 'GPQA Diamond', so it was excluded rather than substituted. LiveCodeBench row is 'v6 (25.02-25.05)'.

Cross-references

This record references no other records.