Llama 3.2 90B Instruct vs Nova Lite
Nova Lite has a slight edge in benchmark performance. Nova Lite is 3.5x cheaper per token.
Meta · Amazon · Updated for 2026
Which is better?
Llama 3.2 90B Instruct outperforms in 3 benchmarks (GPQA, MMLU, MMMU), while Nova Lite is better at 4 benchmarks (ChartQA, DocVQA, MATH, TextVQA). Nova Lite has a slight edge in benchmark performance.
On price, Nova Lite is roughly 3.5x cheaper per token on a blended 3:1 input/output basis, which adds up quickly at production volume.
Nova Lite also accepts a larger context window (300,000 input tokens), making it the stronger choice for long documents and large codebases.
Based on current benchmark, pricing, and model metadata for 2026.
Choose Llama 3.2 90B Instruct
- you need open weights you can self-host or fine-tune
Choose Nova Lite
- you want the strongest raw capability — it leads on 4 of 7 shared benchmarks
- cost matters — it's about 3.5x cheaper per token
- you process long inputs — it offers a 300,000 token context window
- you want the most recent training data — it shipped Nov 2024
At a glance
The differences that matter most.
Performance Benchmarks
Comparative analysis across standard metrics
Llama 3.2 90B Instruct outperforms in 3 benchmarks (GPQA, MMLU, MMMU), while Nova Lite is better at 4 benchmarks (ChartQA, DocVQA, MATH, TextVQA).
Nova Lite has a slight edge in benchmark performance.
Arena Performance
Playground indexes and blind preference scores
Pricing Analysis
Price comparison per million tokens
For input processing, Llama 3.2 90B Instruct ($0.35/1M tokens) is 5.8x more expensive than Nova Lite ($0.06/1M tokens).
For output processing, Llama 3.2 90B Instruct ($0.40/1M tokens) is 1.7x more expensive than Nova Lite ($0.24/1M tokens).
In conclusion, Llama 3.2 90B Instruct is more expensive than Nova Lite.*
* Using a 3:1 ratio of input to output tokens
Context Window
Maximum input and output token capacity
Nova Lite accepts 300,000 input tokens compared to Llama 3.2 90B Instruct's 128,000 tokens. Llama 3.2 90B Instruct can generate longer responses up to 128,000 tokens, while Nova Lite is limited to 2,048 tokens.
Input Capabilities
Supported data types and modalities
Both Llama 3.2 90B Instruct and Nova Lite support multimodal inputs.
They are both capable of processing various types of data, offering versatility in application.
Llama 3.2 90B Instruct
Nova Lite
License
Usage and distribution terms
Llama 3.2 90B Instruct is licensed under Llama 3.2, while Nova Lite uses a proprietary license.
License differences may affect how you can use these models in commercial or open-source projects.
Llama 3.2
Open weights
Proprietary
Closed source
Release Timeline
When each model was launched
Llama 3.2 90B Instruct was released on 2024-09-25, while Nova Lite was released on 2024-11-20.
Nova Lite is 2 months newer than Llama 3.2 90B Instruct.
Sep 25, 2024
1.9 years ago
Nov 20, 2024
1.8 years ago
1mo newerKnowledge Cutoff
When training data ends
Neither model specifies a knowledge cutoff date.
Unable to compare the recency of their training data.
Provider Availability
Llama 3.2 90B Instruct is available from DeepInfra, Bedrock, Fireworks, Together, Hyperbolic. Nova Lite is available from Bedrock.
Llama 3.2 90B Instruct
Nova Lite
Outputs Comparison
Judge for yourself.
Run your own prompts against Llama 3.2 90B Instruct and Nova Lite side-by-side, then vote on the output you prefer.
FAQ
Common questions about Llama 3.2 90B Instruct vs Nova Lite.