BlogLLM Benchmark Scores as Published, Side by Side
BIG-Bench Extra Hard: Results as Published
Every result the catalog holds for BIG-Bench Extra Hard: 6 models from 1 developer, read from 1 page. 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 Google DeepMindIntroduced 2025Data version 2026103001, as of 2026-10-30
BIG-Bench Extra Hard (BBEH) tests general reasoning. Every task of the earlier BIG-Bench Hard is replaced by a new task that targets the same skill at a much higher difficulty, so that current models no longer saturate it. The score is accuracy; the paper reports a harmonic average over tasks.
BIG-Bench Extra Hard
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Gemma 4 12BGoogle, released 2026-04-02 | 53% | Google, Gemma 4 model cardchecked 2026-10-15 | - |
| Gemma 4 26B A4BGoogle, released 2026-04-02 | 64.8% | Google, Gemma 4 model cardchecked 2026-10-15 | - |
| Gemma 4 31BGoogle, released 2026-04-02 | 74.4% | Google, Gemma 4 model cardchecked 2026-10-15 | - |
| Gemma 4 E2BGoogle, released 2026-04-02 | 21.9% | Google, Gemma 4 model cardchecked 2026-10-15 | - |
| Gemma 4 E4BGoogle, released 2026-04-02 | 33.1% | Google, Gemma 4 model cardchecked 2026-10-15 | - |
| Gemma 3 27BGoogle, released 2025-03-12 | 19.3% | Google, Gemma 4 model cardchecked 2026-10-15 | Effort: no think |
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.