BlogLLM Benchmark Scores as Published, Side by Side
MGSM: Results as Published
Every result the catalog holds for MGSM: 7 models from 2 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 Science and MathMaintainer Google (Freda Shi and co-authors)Introduced 2022Data version 2026103001, as of 2026-10-30
MGSM tests multilingual math reasoning with 250 GSM8K word problems translated by hand into ten languages. The score is the share of problems with the correct final answer, averaged over languages.
MGSM
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Llama 4 MaverickMeta, released 2025-04-05 | 92.3% | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| Llama 4 ScoutMeta, released 2025-04-05 | 90.6% | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| Llama 3.3 70BMeta, released 2024-12-06 | 91.1% | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| Llama 3.1 405BMeta, released 2024-07-23 | 91.6% | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| DeepSeek-V4-Flash-BaseDeepSeek, released 2026-04-24 | 85.7% | DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15 | - |
| DeepSeek-V4-Pro-BaseDeepSeek, released 2026-04-24 | 84.4% | DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15 | - |
| DeepSeek-V3.2-BaseDeepSeek, released 2025-12-01 | 81.3% | DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15 | - |
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.