Qwen/Qwen3.5-122B-A10B model card

Fetched 2026-10-18. Weights repo created 2026-02-24 (HF API createdAt 2026-02-24T09:43:37Z, commit dc4d348443bc). Card nameplate: Qwen3.5-122B-A10B, 122B total and 10B activated, 262,144 native context, Apache 2.0, vision-language. The card benchmark table is a vendor claim, not re-run here: MMLU-Pro 86.7, GPQA Diamond 86.6, LiveCodeBench v6 75.6, HMMT Feb 2026 90.8, HMMT Nov 2025 90.2, IFEval 93.2, SWE-bench Verified 72.2, Terminal-Bench 2.0 49.4, MMMU-Pro 75.6, MathVision 83.8. The same table puts Qwen3-235B-A22B-2507 at MMLU-Pro 84.4. Safetensors index total_size is 250,173,007,840 bytes (125.087B BF16 parameters including the vision tower and an MTP block). Per-tensor shapes from the 39 shard headers, fetched the same day, split that into language_model 121,348,687,872, mtp 2,523,679,232, lm_head 762,839,040 and vision 451,290,864. params_b uses the language-model count.