BlogLLM Benchmark Scores as Published, Side by Side

HellaSwag: Results as Published

Every result the catalog holds for HellaSwag: 6 models from 3 developers, read from 2 pages. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.

Area General Knowledge and ReasoningMaintainer University of Washington and Allen Institute for AI (Rowan Zellers and co-authors)Introduced 2019Data version 2026103001, as of 2026-10-30

HellaSwag tests commonsense inference: the model reads the start of an everyday situation and picks the most plausible continuation from several candidates, where the wrong ones were machine-written to fool models. The score is accuracy.

HellaSwag

ModelValuePrinted bySetup as printed
DeepSeek-V4-Flash-BaseDeepSeek, released 2026-04-2485.7%DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15; printed on 1 more page-
DeepSeek-V4-Pro-BaseDeepSeek, released 2026-04-2488%DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15; printed on 1 more page-
DeepSeek-V3.2-BaseDeepSeek, released 2025-12-0186.4%DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15-
Kimi K2 BaseMoonshot AI, released 2025-07-1194.6%Xiaomi, Xiaomi MiMo-V2.5-Pro model cardchecked 2026-10-15Harness: 10-shot
MiMo-V2.5 BaseXiaomi, released 2026-04-2788.6%Xiaomi, Xiaomi MiMo-V2.5-Pro model cardchecked 2026-10-15Harness: 10-shot
MiMo-V2.5-Pro BaseXiaomi, released 2026-04-2789.8%Xiaomi, Xiaomi MiMo-V2.5-Pro model cardchecked 2026-10-15Harness: 10-shot

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.