Skip to content

Llama 3.1 405B vs Mistral Large 2

Meta
Llama 3.1 405B

Meta · 405B params · Quality: 88

Mistral
Mistral Large 2

Mistral AI · 123B params · Quality: 82

Architecture Comparison

SpecLlama 3.1 405BMistral Large 2
TypeDENSEDENSE
Total Parameters405B123B
Active Parameters405B123B
Layers12688
Hidden Dimension16,38412,288
Attention Heads12896
KV Heads88
Context Length131,072131,072
Precision (default)BF16BF16

Memory Requirements

PrecisionLlama 3.1 405BMistral Large 2
BF16 Weights810.0 GB246.0 GB
FP8 Weights405.0 GB123.0 GB
INT4 Weights202.5 GB61.5 GB
KV-Cache / Token516096 B360448 B
Activation Estimate5.00 GB3.50 GB

Minimum GPUs Needed (BF16)

H100 SXMN/A4 GPUs
L40SN/A7 GPUs

Quality Benchmarks

BenchmarkLlama 3.1 405BMistral Large 2
Overall8882
MMLU88.684.0
HumanEval61.053.0
GSM8K96.891.2
MT-Bench88.084.0

Llama 3.1 405B

MMLU
88.6
HumanEval
61.0
GSM8K
96.8
MT-Bench
88.0

Mistral Large 2

MMLU
84.0
HumanEval
53.0
GSM8K
91.2
MT-Bench
84.0

Capabilities

FeatureLlama 3.1 405BMistral Large 2
Tool Use✓ Yes✓ Yes
Vision✗ No✗ No
Code✓ Yes✓ Yes
Math✓ Yes✓ Yes
Reasoning✗ No✗ No
Multilingual✓ Yes✓ Yes
Structured Output✓ Yes✓ Yes

API Pricing Comparison

Cheapest Output (Llama 3.1 405B)

$3.00/M

Input: $3.00/M

Cheapest Output (Mistral Large 2)

$2.50/M

Input: $2.50/M

ProviderLlama 3.1 405B In $/MOut $/MMistral Large 2 In $/MOut $/M
together$3.50$3.50$2.50$2.50
fireworks$3.00$3.00
mistral$2.00$6.00

Recommendation Summary

  • Llama 3.1 405B scores higher on overall quality (88 vs 82).
  • Mistral Large 2 is cheaper per output token ($2.50/M vs $3.00/M).
  • Mistral Large 2 has a smaller memory footprint (246.0 GB vs 810.0 GB BF16), making it easier to deploy on fewer GPUs.
  • Llama 3.1 405B is stronger at code generation (HumanEval: 61.0 vs 53.0).
  • Llama 3.1 405B is better at math reasoning (GSM8K: 96.8 vs 91.2).

Compare Other Models