Compare · ModelsLive · 2 picked · head to head
Llama 3.2 90B vs Llama 3.3 70B Instruct
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Llama 3.3 70B Instruct wins on 4/5 benchmarks
Llama 3.3 70B Instruct wins 4 of 5 shared benchmarks. Leads in knowledge · math.
Category leads
knowledge·Llama 3.3 70B Instructmath·Llama 3.3 70B Instruct
Hype vs Reality
Attention vs performance
Llama 3.2 90B
#217 by perf·#18 by attention
Llama 3.3 70B Instruct
#218 by perf·#18 by attention
Best value
Llama 3.3 70B Instruct
Llama 3.2 90B
n/a
no price
Llama 3.3 70B Instruct
171.9 pts/$
$0.21/M
Vendor risk
Who is behind the model
Meta AI
$1.87T·Tier 1
Meta AI
$1.87T·Tier 1
Head to head
5 benchmarks · 2 models
Llama 3.2 90BLlama 3.3 70B Instruct
Balrog
Llama 3.2 90B leads by +4.3
Balrog · benchmarks AI agents on text-based adventure games, testing language understanding, strategic planning, and long-horizon reasoning.
Llama 3.2 90B
27.3
Llama 3.3 70B Instruct
23.0
GPQA diamond
Llama 3.3 70B Instruct leads by +8.5
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Llama 3.2 90B
21.4
Llama 3.3 70B Instruct
29.9
MATH level 5
Llama 3.3 70B Instruct leads by +2.2
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
Llama 3.2 90B
39.4
Llama 3.3 70B Instruct
41.6
MMLU
Llama 3.3 70B Instruct leads by +8.0
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
Llama 3.2 90B
73.7
Llama 3.3 70B Instruct
81.7
OTIS Mock AIME 2024-2025
Llama 3.3 70B Instruct leads by +2.5
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Llama 3.2 90B
2.5
Llama 3.3 70B Instruct
5.0
Full benchmark table
| Benchmark | Llama 3.2 90B | Llama 3.3 70B Instruct |
|---|---|---|
Balrog Balrog · benchmarks AI agents on text-based adventure games, testing language understanding, strategic planning, and long-horizon reasoning. | 27.3 | 23.0 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 21.4 | 29.9 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 39.4 | 41.6 |
MMLU Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge. | 73.7 | 81.7 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 2.5 | 5.0 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| — | — | — | — | |
| $0.10 | $0.32 | 131K tokens (~66 books) | $1.55 |