Model card
Llama-4-Maverick.
Metaopen-source400B total / 17B active (128 experts) paramsMixture-of-Experts Transformer
Meta Llama 4 Maverick. 128-expert MoE. Released April 2025. Context: 1M tokens.
§ 01 · Benchmarks
Every benchmark Llama-4-Maverick has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | ARC-Challenge | Reasoning · Commonsense Reasoning | accuracy | 97.4% | #3 | — | source ↗ |
| 02 | GSM8K | Reasoning · Mathematical Reasoning | accuracy | 98.7% | #7 | — | source ↗ |
| 03 | MBPP | Computer Code · Code Generation | pass@1 | 77.6% | #12 | — | source ↗ |
| 04 | MMLU | Reasoning · Commonsense Reasoning | accuracy | 89.4% | #18 | — | source ↗ |
| 05 | GPQA | Reasoning · Multi-step Reasoning | accuracy | 69.8% | #19 | — | source ↗ |
| 06 | MATH | Reasoning · Mathematical Reasoning | accuracy | 89.4% | #19 | — | source ↗ |
| 07 | LiveCodeBench | Computer Code · Code Generation | pass@1 | 43.4% | #23 | — | source ↗ |
| 08 | PLCC | Natural Language Processing · Polish Cultural Competency | history | 76.0% | #73 | — | source ↗ |
| 09 | PLCC | Natural Language Processing · Polish Cultural Competency | grammar | 59.0% | #82 | — | source ↗ |
| 10 | PLCC | Natural Language Processing · Polish Cultural Competency | geography | 71.0% | #88 | — | source ↗ |
| 11 | PLCC | Natural Language Processing · Polish Cultural Competency | average | 58.2% | #95 | — | source ↗ |
| 12 | PLCC | Natural Language Processing · Polish Cultural Competency | art-and-entertainment | 46.0% | #97 | — | source ↗ |
| 13 | PLCC | Natural Language Processing · Polish Cultural Competency | culture-and-tradition | 52.0% | #102 | — | source ↗ |
| 14 | PLCC | Natural Language Processing · Polish Cultural Competency | vocabulary | 45.0% | #105 | — | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 02 · Strengths by area
Where Llama-4-Maverick actually performs.
§ 04 · Related models
Other Meta models scored on Codesota.
§ 05 · Sources & freshness
Where these numbers come from.
sdadas/PLCC
7
results
meta-blog
5
results
meta-model-card
2
results
14 of 14 rows marked verified.