Mixtral 8x22B Specs, Benchmarks & Downloads
Mixtral 8x22B
Larger mixture of experts model
Status
active
Parameters
141.00
Context Window
65536
Modality
Text
Release Date
Apr 17, 2024
About
Explore Mixtral 8x22B, a text AI model developed by Mistral AI. View specifications, context window, architecture, license, supported platforms, features, benchmarks, official downloads, and related articles.
Mixtral 8x22B provides stronger reasoning through sparse expert routing.
Family
MixtralDeveloper
Mistral AIArchitecture
Mixture of ExpertsLicense
Mistral LicenseHardware Requirements
Minimum:
Use Case: Local inference
GPU Minimum: 2x RTX 4090
GPU VRAM: 48GB
System RAM: 128GB
Storage: 150GB
Operating System: Linux
Mixtral 8x22B quantized deployment
Recommended:
Use Case: Server inference
GPU Minimum: 4x RTX 4090
GPU VRAM: 96GB
System RAM: 256GB
Storage: 250GB
Operating System: Linux
Large MoE deployment
High Performance:
Use Case: Production serving
GPU Minimum: 4x H100
GPU VRAM: 320GB
System RAM: 512GB
Storage: 500GB
Operating System: Linux
Enterprise MoE serving
Model Deployment Readiness
Model Ecosystem Support
18%
Platforms
33%
Features
1%
Downloads
20%
Platform Compatibility
33%
Supported across 4 of 12 platforms
Context Window
Industry-standard context window comparison
65,536
tokens
64K
4K
32K
128K
1M
2M
Database Rank
#28
Percentile
Top 73%
Capacity
High
Benchmark Performance
Evaluation profile across publicly reported benchmarks
|
HumanEval
|
Coding | 52.2 % | 2024 |
|
MMLU
|
Knowledge | 81.0 % | 2024 |
|
GPQA
|
Reasoning | 41.0 % | 2024 |
|
HumanEval
|
Coding | 60.0 % | 2024 |
|
GSM8K
|
Mathematics | 80.0 % | 2024 |
|
HellaSwag
|
Reasoning | 87.0 % | 2024 |
|
MMLU
|
Knowledge | 77.8 % | 2024 |
|
MATH-500
|
Mathematics | 78.0 % | 2024 |
Highest Benchmark
HellaSwag
87.0%
Benchmarks Available
8
Latest Evaluation
Apr 2024
