BlogLLM Benchmark Scores as Published, Side by Side
Llama 4 Maverick: Benchmark Scores as Published
Every result the catalog holds for Llama 4 Maverick: 22 tests in 8 areas, from 2 pages. 13 printed by Meta itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.
Developer MetaReleased 2025-04-05Context window 1MInput Text, imageOutput TextWeights Open (Llama 4 Community License Agreement)Data version 2026103001, as of 2026-10-30
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 69.8% | Meta, Meta Llama 4 model cardchecked 2026-10-15; printed on 1 more page | - |
| MATH-500 | 90% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| RULER at 128K | 86.7% | Mistral AI, Medium is the new largechecked 2026-10-15 | Context: 128000 |
| RULER at 32K | 94.8% | Mistral AI, Medium is the new largechecked 2026-10-15 | Context: 32000 |
| MTOB (full book) eng->kgv (chrF) | 50.8 | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| MTOB (full book) kgv->eng (chrF) | 46.7 | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| MTOB (half book) eng->kgv (chrF) | 54 | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| MTOB (half book) kgv->eng (chrF) | 46.4 | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
Multimodal Understanding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMMU-Pro | 59.6% | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| MMMU (validation) | 73.4% | Meta, Meta Llama 4 model cardchecked 2026-10-15; printed on 1 more page | - |
| MathVista (testmini) | 73.7% | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| AI2D (test) | 84.4% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| ChartQA | 90% | Meta, Meta Llama 4 model cardchecked 2026-10-15; printed on 1 more page | - |
| DocVQA (ANLS) | 94.4 | Meta, Meta Llama 4 model cardchecked 2026-10-15; printed on 1 more page | - |
General Knowledge and Reasoning
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMLU-Pro | 80.5% | Meta, Meta Llama 4 model cardchecked 2026-10-15; printed on 1 more page | - |
Code Generation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LiveCodeBench v6 | 28.7% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| LiveCodeBench (version not stated) | 43.4% | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
| MultiPL-E | 76.4% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| HumanEval | 85.4% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
Instruction Following
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| IFEval | 88.9% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
Multilingual
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MGSM | 92.3% | Meta, Meta Llama 4 model cardchecked 2026-10-15 | - |
Preference and Composite Rankings
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Arena-Hard win rate (GPT-4-1106 judge) | 91.8% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
More from Meta
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the 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.