GPT-5 Codex vs Llama 3.2 90B Instruct
OpenAI vs Meta — benchmarks, pricing, and capabilities side by side.
- •GPT-5 Codex has the higher intelligence index (87.1 vs 54)
- •Llama 3.2 90B Instruct is cheaper ($0.35 vs $1.25 per 1M input)
- •GPT-5 Codex is faster
- •GPT-5 Codex has a larger context window (400K)
| GPT-5 Codex | Llama 3.2 90B Instruct | |
|---|---|---|
| Intelligence index | 87.1 | 54 |
| Developer | OpenAI | Meta |
| Type | Multimodal | Multimodal |
| Access | API only | Open weights |
| Context window | 400,000 tokens | 128,000 tokens |
| Input price | $1.25 / 1M | $0.35 / 1M |
| Output price | $10.00 / 1M | $0.40 / 1M |
| Speed | 180 tok/s | 100 tok/s |
| Released | September 23, 2025 | September 25, 2024 |
| Parameters | — | 90000000000 |
| Input modalities | Text, Image | — |
| Output modalities | Text | — |
Shared benchmarks
GPT-5 Codex
Llama 3.2 90B Instruct
GPQA Diamond
83.7
46.7
Humanity’s Last Exam
25.6
4.9
LiveCodeBench
84
21.4
MMLU-Pro
86.5
67.1