Llama 3.2 Instruct 90B (Vision)
Intelligence
11.9
#414/523
Coding
—
Math
—
Speed
41 tok/s
TTFT: 392.00s
Pricing
$0.72 / $0.72
per 1M tokens (in/out)
Llama 3.2 Instruct 90B (Vision) is Meta’s latest model designed for advanced instruction-following tasks. It processes at 41.137 tokens per second and is priced at $0.72 per million tokens, targeting professional users in fields requiring sophisticated AI interactions.
When to Use Llama 3.2 Instruct 90B (Vision)
✓ Best For
- Instruction-based applications
- Vision-related tasks
- Professional content generation
✗ Not Ideal For
- Casual users
- Basic coding or math tasks
How Llama 3.2 Instruct 90B (Vision) Compares
Intelligence Index · Higher is better
Reka AIAlibabaMetaUpstage
Benchmark Profile
Output Speed · tok/s
DeepSeekAlibabaMetaOpenAIZ AI
Intelligence · Coding · Math
Intelligence
Coding
Math
All Benchmark Scores (8)
| Benchmark | Score | |
|---|---|---|
| Intelligence Index | 11.9 | |
| MMLU-Pro | 671% | |
| GPQA | 432% | |
| LiveCodeBench | 214% | |
| HLE | 49% | |
| SciCode | 24% | |
| AIME | 5% | |
| MATH 500 | 62.9% |
Data: Artificial Analysis · Updated: April 9, 2026
Frequently Asked Questions (15)
When was Llama 3.2 Instruct 90B (Vision) released?
Llama 3.2 Instruct 90B (Vision) was released on September 25, 2024.
Who created Llama 3.2 Instruct 90B (Vision)?
Llama 3.2 Instruct 90B (Vision) was created by Meta.
How intelligent is Llama 3.2 Instruct 90B (Vision)?
Llama 3.2 Instruct 90B (Vision) scores 12 (estimated) on the Artificial Analysis Intelligence Index, placing it below average among other open weight non-reasoning models of similar size (median: 13).
How fast is Llama 3.2 Instruct 90B (Vision)?
Llama 3.2 Instruct 90B (Vision) generates output at 54.8 tokens per second (based on the median across providers serving the model), which is below average compared to other open weight non-reasoning models of similar size (median: 59.9 t/s).
What is the latency of Llama 3.2 Instruct 90B (Vision)?
Llama 3.2 Instruct 90B (Vision) has a time to first token (TTFT) of 1.06s (based on the median across providers serving the model), which is very competitive compared to other open weight non-reasoning models of similar size (median: 1.90s).
How much does Llama 3.2 Instruct 90B (Vision) cost?
Llama 3.2 Instruct 90B (Vision) costs $0.72 per 1M input tokens (somewhat higher than average, median: $0.54) and $0.72 per 1M output tokens (better than average, median: $0.90), based on the median across providers serving the model.
What is Llama 3.2 Instruct 90B (Vision) API pricing?
Llama 3.2 Instruct 90B (Vision) costs $0.72 per 1M input tokens and $0.72 per 1M output tokens (based on the median across providers serving the model). For a blended rate (3:1 input to output ratio), this is $0.72 per 1M tokens. Pricing may vary by provider.
Is Llama 3.2 Instruct 90B (Vision) a reasoning model?
No, Llama 3.2 Instruct 90B (Vision) is not a reasoning model. It provides direct responses without extended chain-of-thought reasoning.
What input modalities does Llama 3.2 Instruct 90B (Vision) support?
Llama 3.2 Instruct 90B (Vision) supports image input.
What output modalities does Llama 3.2 Instruct 90B (Vision) support?
Llama 3.2 Instruct 90B (Vision) supports text only output.
Can Llama 3.2 Instruct 90B (Vision) process images?
Yes, Llama 3.2 Instruct 90B (Vision) supports image input and can analyze, describe, and answer questions about images.
Is Llama 3.2 Instruct 90B (Vision) multimodal?
No, Llama 3.2 Instruct 90B (Vision) is not multimodal. It only supports image input.
What is the context window of Llama 3.2 Instruct 90B (Vision)?
Llama 3.2 Instruct 90B (Vision) has a context window of 130k tokens. This determines how much text and conversation history the model can process in a single request.
Is Llama 3.2 Instruct 90B (Vision) open source?
Yes, Llama 3.2 Instruct 90B (Vision) is open weights. The model weights are publicly available and can be downloaded for self-hosting.
How many parameters does Llama 3.2 Instruct 90B (Vision) have?
Llama 3.2 Instruct 90B (Vision) has 90 billion parameters.
