Model card
Llama 3.2 Vision 90B.
Metaopen-sourceUnknown paramsLlama 3.1 + cross-attention vision adapter
90B parameter vision-language model from Llama 3.2 family. September 2024. Source: Llama 3 paper arxiv:2407.21783.
§ 02 · Benchmarks
Every benchmark Llama 3.2 Vision 90B has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | TextVQA | Multimodal · Visual Question Answering | accuracy | 83.4% | #10 | 2024-07-31 | source ↗ |
| 02 | MMMU | Multimodal · Visual Question Answering | accuracy | 60.3% | #22 | 2024-07-31 | 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.
§ 03 · Strengths by area
Where Llama 3.2 Vision 90B actually performs.
§ 04 · Papers
1 paper with results for Llama 3.2 Vision 90B.
- 2024-07-31· Natural Language Processing· 2 results
The Llama 3 Herd of Models
§ 05 · Related models
Other Meta models scored on Codesota.
Llama 4 Maverick
400B total / 17B active (128 experts) params · 15 results
Llama 3.1 405B
13 results
Llama 3 70B
11 results
Llama 3 (405B, Instruct)
9 results
Llama-3.2-1B-Instruct
9 results
Llama-3.2-3B-Instruct
9 results
Llama-4-Scout
109B total / 17B active (16 experts) params · 9 results
Meta-Llama-3.1-405B-Instruct
9 results
§ 06 · Sources & freshness
Where these numbers come from.
arxiv
2
results
2 of 2 rows marked verified.