Proprietary • Mistral AI

Mistral Large 2: Complete Guide

🎯 Overview: Mistral Large 2 is a highly efficient 123B frontier model with exceptional reasoning, function calling, and cost-effectiveness. Available via Mistral API, Azure, and open-weight variants.

Key Specifications

123B Parameters

32K Context Tokens

$0.27 Input Cost/MTok

3x faster Than Large v1

Competitive Advantages

✓ Exceptional Efficiency: 123B parameters achieves frontier performance with lower latency and cost than competitors. 3x faster than Large v1.

✓ Best Reasoning: Top scores on MMLU (84.4%), Math problems (89%), and code benchmarks (HUMANEVAL 91%). Specialized for complex tasks.

✓ Function Calling: Native support for precise tool use. Reliable function argument generation for API integration.

✓ Multiple Deployment Options: Mistral API, Azure Mistral integration, or self-hosted open weights.

Pricing & Plans

Mistral API
Input: $0.27 per million tokens

Output: $0.81 per million tokens

Batch (50% discount): $0.135 input / $0.405 output

Azure Deployments
Pay-as-you-go: Variable pricing based on region

Commitment plans: 25-50% savings with prepaid tokens

Cost Comparison (1M tokens/month)
Mistral Large 2: $270 input + $810 output = $1,080/month

Claude 3.5 Sonnet: $3,000 input + $15,000 output = $18,000/month

Savings: ~95% cheaper than Claude for comparable reasoning

Performance Benchmarks

Benchmark	Mistral Large 2	Llama 3.1 70B	GPT-4o
MMLU (Knowledge)	84.4%	85.2%	92.3%
GSM-8K (Math)	89.0%	93.0%	95.1%
HUMANEVAL (Code)	91.0%	85.9%	92.3%
Speed (tokens/sec)	Very Fast	Fast	Very Fast
Efficiency (params/performance)	Best	Good	Excellent

Best Use Cases

🛠️ API & Tool Integration
Function calling enables reliable interaction with external systems, APIs, and databases

💡 Complex Problem Solving
Mathematics, logic puzzles, multi-step reasoning tasks benefit from strong reasoning capabilities

💰 Cost-Optimized Enterprise
Frontier-quality reasoning at 95% lower cost than competitors. Ideal for high-volume deployments

⚡ Low-Latency Systems
Fast inference speed makes it suitable for real-time applications and high-throughput systems

API Integration Example

from mistralai import Mistral

client = Mistral(api_key="your-api-key")

response = client.chat.complete(

    model="mistral-large-latest",

    messages=[{

        "role": "user",

        "content": "Solve this math problem: 5x + 3 = 18"

    }],

    tools=[{

        "type": "function",

        "function": {

            "name": "calculator",

            "description": "Performs mathematical operations"

        }

    }]

)

print(response.choices[0].message.content)

Comparison: Mistral vs Alternatives

Feature	Mistral Large 2	Claude 3.5 Sonnet	GPT-4o	Llama 3.1 70B
Cost/MTok	$0.27	$3	$5	Free (self-hosted)
Context Window	32K	200K	128K	128K
Reasoning	Strong	Best	Strong	Strong
Function Calling	Native	Native	Native	Can add via LoRA
Inference Speed	Very Fast	Very Fast	Very Fast	Fast

← Back to All Models