Activefrontierllm Open Source

Llama 3.2 90B

by Meta· Released September 2024· Cutoff August 2024

Llama 3.2 90B is Meta's largest multimodal model in the Llama 3.2 family, supporting text and image inputs. It excels at complex reasoning, multilingual tasks, and vision-language understanding, positioning it as a top-tier open-weight model for enterprise and research applications.

Official Site API Docs 🤗 Hugging Face

Input cost

Free (open source)

Output cost

Free (open source)

Context window

128K tokens

Max output

8192 tokens

Modalities

textimage

Parameters

90B

License

Llama 3.2 Community License

Capabilities

Multimodal (text + image input)MultilingualCode GenerationReasoningFunction CallingJSON ModeStreaming

Best For

Complex multimodal reasoning tasks requiring both text and image understanding at scale.

Strengths

State-of-the-art performance on multimodal benchmarks
Large 128K context window
Strong multilingual capabilities
Open-weight with permissive license
Efficient inference for its size

Limitations

Requires significant computational resources for deployment
No image generation capability (input only)
May produce biased or unsafe outputs without fine-tuning
Not as fast as smaller models in the family

Use Cases

Visual question answering

Document understanding and analysis

Multilingual content generation

Complex code generation and debugging

Scientific research and data analysis

Enterprise knowledge retrieval with images

Accessibility tools for visually impaired

Improvements Over Previous Model

First Llama model to support multimodal (image) input
Context window increased from 8K (Llama 3.1 70B) to 128K tokens
Significantly improved multilingual performance
Added vision-language capabilities without sacrificing text-only performance
Better reasoning and coding benchmarks compared to Llama 3.1 70B
More efficient architecture for longer sequences

Back to all models

Activefrontierllm Open Source

Llama 3.2 90B

by Meta· Released September 2024· Cutoff August 2024

Official Site API Docs 🤗 Hugging Face

Input cost

Free (open source)

Output cost

Free (open source)

Context window

128K tokens

Max output

8192 tokens

Modalities

textimage

Parameters

90B

License

Llama 3.2 Community License

Capabilities

Multimodal (text + image input)MultilingualCode GenerationReasoningFunction CallingJSON ModeStreaming

Best For

Complex multimodal reasoning tasks requiring both text and image understanding at scale.

Strengths

State-of-the-art performance on multimodal benchmarks
Large 128K context window
Strong multilingual capabilities
Open-weight with permissive license
Efficient inference for its size

Limitations

Requires significant computational resources for deployment
No image generation capability (input only)
May produce biased or unsafe outputs without fine-tuning
Not as fast as smaller models in the family

Use Cases

Visual question answering

Document understanding and analysis

Multilingual content generation

Complex code generation and debugging

Scientific research and data analysis

Enterprise knowledge retrieval with images

Accessibility tools for visually impaired

Improvements Over Previous Model

First Llama model to support multimodal (image) input
Context window increased from 8K (Llama 3.1 70B) to 128K tokens
Significantly improved multilingual performance
Added vision-language capabilities without sacrificing text-only performance
Better reasoning and coding benchmarks compared to Llama 3.1 70B
More efficient architecture for longer sequences

Back to all models