Compare · ModelsLive · 2 picked · head to head
Llama 3.3 70B Instruct vs Llama 4 Maverick
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Llama 4 Maverick wins on 7/8 benchmarks
Llama 4 Maverick wins 7 of 8 shared benchmarks. Leads in general · knowledge · math.
Category leads
general·Llama 4 Maverickknowledge·Llama 4 Maverickmath·Llama 4 Maverickreasoning·Llama 4 Maverickcoding·Llama 4 Maverick
Head to head · from the data · data as of 2026-11-15
- Benchmarks · Llama 3.3 70B Instruct leads on 1 of 8 shared benchmarks (Lmca), Llama 4 Maverick on 7 (Dtbench, Fiction.LiveBench, GPQA diamond, MATH level 5, 3 more)
- Price · Llama 3.3 70B Instruct input $0.71/M vs Llama 4 Maverick $0.35/M, Llama 4 Maverick is 51% cheaper (OpenRouter list price, as of 2026-11-15)
- Cheapest provider · Llama 3.3 70B Instruct at DeepInfra $0.10/M input (of 9 offers) as of 2026-11-15 · Llama 4 Maverick at Google and 1 other provider $0.35/M input (of 2 offers) as of 2026-11-15
- Context · Llama 3.3 70B Instruct 131K tokens vs Llama 4 Maverick 524K tokens
- Released · Llama 3.3 70B Instruct Dec 6, 2024 vs Llama 4 Maverick Apr 5, 2025 · 4 months apart
- Open weights · Llama 3.3 70B Instruct and Llama 4 Maverick both have open weights
Sources · scores from public leaderboards, list prices and provider offers from OpenRouter, price moves from the BenchGecko changelog (5% or more, held in the next daily check), Gecko tests run by BenchGecko.
Hype vs Reality
Attention vs performance
Llama 3.3 70B Instruct
#230 by perf·#17 by attention
Llama 4 Maverick
#277 by perf·#17 by attention
Best value
Llama 3.3 70B Instruct
1.2x better value than Llama 4 Maverick
Llama 3.3 70B Instruct
51.0 pts/$
$0.71/M
Llama 4 Maverick
41.3 pts/$
$0.68/M
Vendor risk
Who is behind the model
Meta AI
$1.65T·Tier 1
Meta AI
$1.65T·Tier 1
Head to head
8 benchmarks · 2 models
Llama 3.3 70B InstructLlama 4 Maverick
Dtbench
Llama 4 Maverick leads by +4.0
Llama 3.3 70B Instruct
32.5
Llama 4 Maverick
36.5
Fiction.LiveBench
Llama 4 Maverick leads by +12.9
Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination.
Llama 3.3 70B Instruct
33.3
Llama 4 Maverick
46.2
GPQA diamond
Llama 4 Maverick leads by +24.8
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Llama 3.3 70B Instruct
31.2
Llama 4 Maverick
56.0
Lmca
Llama 3.3 70B Instruct leads by +1.9
Llama 3.3 70B Instruct
20.6
Llama 4 Maverick
18.7
MATH level 5
Llama 4 Maverick leads by +31.4
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
Llama 3.3 70B Instruct
41.6
Llama 4 Maverick
73.0
OTIS Mock AIME 2024-2025
Llama 4 Maverick leads by +15.4
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Llama 3.3 70B Instruct
5.0
Llama 4 Maverick
20.5
SimpleBench
Llama 4 Maverick leads by +9.4
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
Llama 3.3 70B Instruct
3.9
Llama 4 Maverick
13.2
WeirdML
Llama 4 Maverick leads by +10.0
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Llama 3.3 70B Instruct
14.4
Llama 4 Maverick
24.5
Full benchmark table
| Benchmark | Llama 3.3 70B Instruct | Llama 4 Maverick |
|---|---|---|
Dtbench | 32.5 | 36.5 |
Fiction.LiveBench Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination. | 33.3 | 46.2 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 31.2 | 56.0 |
Lmca | 20.6 | 18.7 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 41.6 | 73.0 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 5.0 | 20.5 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 3.9 | 13.2 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 14.4 | 24.5 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.71 | $0.71 | 131K tokens (~66 books) | $7.10 | |
| $0.35 | $1.00 | 524K tokens (~262 books) | $5.13 |
Llama 3.3 70B Instruct vs Llama 4 Maverick · questions answered from the data
Llama 3.3 70B Instruct leads on 1 of 8 shared benchmarks (Lmca), Llama 4 Maverick on 7 (Dtbench, Fiction.LiveBench, GPQA diamond, MATH level 5, 3 more)