Mistral's 123B flagship with 128k context, enterprise function calling, and deep multilingual reasoning at $2.00 / $6.00 per 1M tokens.
Input / 1M Tokens
$2.00
20% cheaper than GPT-4o
Output / 1M Tokens
$6.00
40% cheaper than GPT-4o
Context Window
128,000
Tokens native support
MMLU Benchmark
84.0%
Top-tier reasoning
🧮 Live Mistral Large 2 Cost Calculator
Calculate your workload costs and compare them directly against GPT-4o and Claude 3.5 Sonnet.
Input Tokens (Millions / month)10M tokens
Output Tokens (Millions / month)5M tokens
Mistral Large 2 Cost
$50.00
GPT-4o ($2.50 / $10.00)
$75.00
Claude 3.5 Sonnet ($3 / $15)
$105.00
Monthly Savings vs Claude
52.4%
📊 Enterprise Frontier Model Comparison
How Mistral Large 2 stacks up against market leaders across core metrics:
Model
Input / 1M
Output / 1M
MMLU
Context
EU Sovereignty / GDPR
Mistral Large 2
$2.00
$6.00
84.0%
128k
Native Paris / EU Data Centers
GPT-4o
$2.50
$10.00
88.7%
128k
US / Global Cloud
Claude 3.5 Sonnet
$3.00
$15.00
88.3%
200k
US Hosted (AWS / GCP)
Llama 3.3 70B
$0.59
$0.79
86.0%
128k
Open-Weights Multi-Host
🚀 Where to Deploy Mistral Large 2
Mistral Large 2 is available across three primary enterprise deployment routes:
Mistral La Plateforme: Native direct API with earliest updates, function calling studio, fine-tuning, and direct European data residency.
Azure AI Foundry: Pay-as-you-go serverless endpoint or dedicated provisioned throughput with Microsoft enterprise agreement billing.
AWS Bedrock: Seamlessly integrate into existing Amazon VPC environments with zero external egress.
Frequently Asked Questions
How much does Mistral Large 2 cost via API?
Mistral Large 2 (mistral-large-2407) is priced at $2.00 per 1 million input tokens and $6.00 per 1 million output tokens on Mistral's La Plateforme, Azure AI Foundry, and AWS Bedrock.
How does Mistral Large 2 compare to GPT-4o?
Mistral Large 2 scores 84.0% on MMLU compared to 88.7% for GPT-4o. It is 20% cheaper on input tokens ($2.00 vs $2.50) and 40% cheaper on output tokens ($6.00 vs $10.00), offering superior native performance in European languages and strict GDPR compliance.
Can you self-host Mistral Large 2?
Yes, Mistral Large 2 has 123 billion parameters and open model weights under the Mistral Research License. In FP8 quantization, it requires approximately 135GB VRAM, running on 2x NVIDIA A100 (80GB) or 2x H100 GPUs.