Meta model product
Llama 3.2 90B
2 90B is a large multimodal language model optimized for visual recognition, image reasoning, and captioning tasks.
Updated Aug 17, 2026. Default version: Llama 3.2 90B Instruct
LLMBoard Score9.9Llama 3.2 90B Instruct
Coverage60%12 benchmark families
Context window128KTokens
Official input priceN/AOfficial price unavailable
Llama 3.2 90B Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Llama 3.2 90B Benchmark Results
Benchmark scores for Llama 3.2 90B Instruct.
Sort by
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkInfographicsQA | Score56.8% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkVQAv2 | Score78.1% | Rank03 | Participants3 | Percentile0.0% | EvidenceC | Evaluated |
| BenchmarkAI2D | Score92.3% | Rank10 | Participants33 | Percentile71.9% | EvidenceC | Evaluated |
| BenchmarkTextVQA | Score73.5% | Rank12 | Participants16 | Percentile26.7% | EvidenceC | Evaluated |
| BenchmarkChartQA | Score85.5% | Rank13 | Participants26 | Percentile52.0% | EvidenceC | Evaluated |
| BenchmarkMGSM | Score86.9% | Rank14 | Participants31 | Percentile56.7% | EvidenceC | Evaluated |
| BenchmarkDocVQA | Score90.1% | Rank21 | Participants28 | Percentile25.9% | EvidenceC | Evaluated |
| BenchmarkMathVista | Score57.3% | Rank29 | Participants39 | Percentile26.3% | EvidenceC | Evaluated |
| BenchmarkMMLU | Score86.0% | Rank33 | Participants101 | Percentile68.0% | EvidenceC | Evaluated |
| BenchmarkMATH | Score68.0% | Rank44 | Participants71 | Percentile38.6% | EvidenceC | Evaluated |
| BenchmarkMMMU | Score60.3% | Rank44 | Participants63 | Percentile30.6% | EvidenceC | Evaluated |
| BenchmarkMMMU-Pro | Score45.2% | Rank62 | Participants68 | Percentile9.0% | EvidenceC | Evaluated |
| BenchmarkGPQA | Score46.7% | Rank192 | Participants239 | Percentile19.8% | EvidenceC | Evaluated |
Llama 3.2 90B Arena Results
Preference and agent-evaluation results for the default version.
Sort by
Arena | Category | Rank | Rating / score | Votes | Observations | Result date |
|---|
| Arenavision style control | Categoryoverall | Rank126 | Rating / score1032.7 | Votes8,682 | ObservationsN/A | Result date |
| Arenavision | Categoryoverall | Rank127 | Rating / score999.8 | Votes8,682 | ObservationsN/A | Result date |
Llama 3.2 90B Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- N/A
- Tracked offerings
- 0
No provider prices
The default version has no current input or output token prices.
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Llama 3.2 90B Runtime Performance
Provider-specific output speed and catalog latency for Llama 3.2 90B Instruct. Runtime does not affect the capability score.
Sort by
Provider | Output Speed | Catalog Latency | Max Input | Max Output | Updated |
|---|
| ProviderBedrock | Output Speed100 tok/s | Catalog Latency0.5 s | Max Input128K | Max Output128K | Updated |
| ProviderTogether | Output Speed57 tok/s | Catalog Latency0.5 s | Max Input128K | Max Output128K | Updated |
| ProviderFireworks | Output Speed50 tok/s | Catalog Latency0.5 s | Max Input128K | Max Output128K | Updated |
| ProviderHyperbolic | Output Speed42 tok/s | Catalog Latency0.5 s | Max Input128K | Max Output128K | Updated |
| ProviderDeepInfra | Output Speed24 tok/s | Catalog Latency0.5 s | Max Input128K | Max Output128K | Updated |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Llama 3.2 90B Specifications
Technical details for the model's default version.
- Version
- Llama 3.2 90B Instruct
- Released
- Sep 25, 2024
- Knowledge cutoff
- Unknown
- Parameters
- 90B
- Context window
- 128K
- Max output
- 128K
- Inputs
- image, text
- Outputs
- text
- Open weights
- No
- License
- Llama 3.2
Llama 3.2 90B Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
Sort by
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionLlama 3.2 90B Instruct | Released | LLMBoard9.9 | Parameters90B | Context128K | Max output128K | Open weightsNo | LicenseLlama 3.2 |
Llama 3.2 90B vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to Llama 3.2 90B
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
ME
Llama 3.1 70B
Meta
ME
Llama 4 Scout
Meta
ME
Llama 3.2 3B
Meta
ME
Llama 3.2 11B
Meta
ME
Llama 3.1 8B
Meta
ME
Llama 3.3 70B
Meta
What is Llama 3.2 90B?
Key information about Llama 3.2 90B and its available data.
2 90B is a large multimodal language model optimized for visual recognition, image reasoning, and captioning tasks. It supports a context length of 128,000 tokens and is designed for deployment on edge and mobile devices, offering state-of-the-art performance in image understanding and generative tasks.
Data as of 2026-08-17.
FAQ
Common questions about Llama 3.2 90B.
When was Llama 3.2 90B released?
Llama 3.2 90B's default version was released on Sep 25, 2024.
How much does Llama 3.2 90B cost?
No official standard PAYG price is currently available for Llama 3.2 90B.
Who created Llama 3.2 90B?
Llama 3.2 90B was created by Meta.
What is the context window for Llama 3.2 90B?
The default version has a 128K token context window.
Is Llama 3.2 90B open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer Llama 3.2 90B?
No provider offering is currently linked to the default version.
What models should I compare Llama 3.2 90B with?
Nearby ranked alternatives include Gemini 1.5 Flash, Gemma 3 12B, Qwen2.5 Omni 7B.