Compare/GPT-4.5 (Preview) vs Llama 3.2 Instruct 90B (Vision)

GPT-4.5 (Preview)vsLlama 3.2 Instruct 90B (Vision)

Side-by-side comparison of pricing, 12 benchmarks, and generation speed.

OpenAI

GPT-4.5 (Preview)

Input
$0/M
Output
$0/M
Speed
TTFT
Meta

Llama 3.2 Instruct 90B (Vision)

Input
$0.72/M
Output
$0.72/M
Speed
42 tok/s
TTFT
0.37s

Winner by Category

Cheaper
GPT-4.5 (Preview)
Faster (tok/s)
Llama 3.2 Instruct 90B (Vision)
Lower Latency
GPT-4.5 (Preview)
Benchmarks (1-6)
Llama 3.2 Instruct 90B (Vision)

Pricing Comparison

MetricGPT-4.5 (Preview)Llama 3.2 Instruct 90B (Vision)
Input ($/M tokens)$0$0.72
Output ($/M tokens)$0$0.72
Cost for 1M input + 100K output tokens:
GPT-4.5 (Preview)$0.00
Llama 3.2 Instruct 90B (Vision)$0.79

Speed Comparison

Output Speed (tokens/s) — higher is better
GPT-4.5 (Preview)
Llama 3.2 Instruct 90B (Vision)
42 tok/s
Time to First Token (seconds) — lower is better
GPT-4.5 (Preview)
Llama 3.2 Instruct 90B (Vision)
0.37s

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
20.011.9
Coding Index
Math Index
GPQA Diamond
43.2%
MMLU-Pro
67.1%
LiveCodeBench
21.4%
AIME 2025
MATH-500
62.9%
Humanity's Last Exam
4.9%
SciCode
24.0%
IFBench
TerminalBench
GPT-4.5 (Preview)1 wins
6 winsLlama 3.2 Instruct 90B (Vision)

Frequently Asked Questions

Which is cheaper, GPT-4.5 (Preview) or Llama 3.2 Instruct 90B (Vision)?

GPT-4.5 (Preview) is cheaper overall. Its blended price (3:1 input/output ratio) is $0.00/M tokens vs $0.72/M for Llama 3.2 Instruct 90B (Vision).

Which model performs better on benchmarks?

Llama 3.2 Instruct 90B (Vision) wins 6 out of 12 benchmarks compared to 1 for GPT-4.5 (Preview). See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

Llama 3.2 Instruct 90B (Vision) generates tokens faster at 42 tok/s vs 0 tok/s. GPT-4.5 (Preview) also has lower time-to-first-token (0.00s vs 0.37s).

When should I use GPT-4.5 (Preview) vs Llama 3.2 Instruct 90B (Vision)?

Choose based on your priorities: GPT-4.5 (Preview) for lower cost, Llama 3.2 Instruct 90B (Vision) for stronger benchmark performance, and Llama 3.2 Instruct 90B (Vision) for faster generation. For latency-sensitive apps, check the TTFT comparison above.