CheckAPICost 2026 INDEX

Mistral AI • Enterprise Instruction

Mistral Small 3 API Pricing, Vision Benchmarks & Rate Card (2026)

Complete 2026 developer pricing guide for Mistral Small 3. Benchmark token economics, context window limits, prompt caching discounts, and live inference latency across official provider endpoints.

⚡ 60-Second Mistral Small Verdict ✓ Verified October 2026

24B Multimodal Vision • 140+ Tokens/Sec

Mistral Small 3 Pricing & Vision Verdict: Mistral Small 3 (24B parameters) combines enterprise-grade multimodal vision and low-latency reasoning at $0.20 per 1M input tokens and $0.60 per 1M output tokens. Operating at over 140 tokens/second, it outperforms GPT-4o-mini on math and document reasoning while fitting on a single 24GB RTX 4090 or A10G GPU for self-hosting. Explore models in our Models Directory and compare vision tools in Image & Video Hub.

Standard Input Rate $0.20 / 1M Tokens

Standard Output Rate $0.60 / 1M Tokens

Inference Throughput 140+ Tokens / Second

Self-Hosting Hardware 1x RTX 4090 or A10G (24GB)

Input Token Rate

$0.20

per 1M input tokens

Output Token Rate

$0.60

per 1M generated tokens

Context Window

32,768

maximum token buffer

Avg Throughput

110

tokens/sec (typical speed)

Billing Dimension	Standard Rate	Cached / Optimized Rate	Batch (50% Off)	Context Window
Input Prompt	$0.20 / M	$0.10 / M	50% discount (24h SLA)	32,768 tokens
Output Generation	$0.60 / M	Standard metered	50% discount (24h SLA)	Up to max model limit
Tool / Function Calling	Standard input tokens	Eligible for prefix cache	Batch supported	Shared context budget

Interactive Monthly Cost Estimator

Estimate your total bill based on prompt size, response volume, and request volume.

Monthly Requests

Avg Input Tokens per Request

Avg Output Tokens per Request

Prompt Cache Hit Ratio (%)

Estimated Monthly Bill

$0.00

Calculated using live 2026 rates for Mistral Small 3

Architectural Breakdown & Cost Optimization

Mistral Small 3 represents a compelling value proposition in the 2026 developer ecosystem, balancing high accuracy with predictable operating margins. When designing agentic workflows or production RAG systems around Mistral Small 3, prompt engineering and caching strategy represent the single largest levers on your monthly bill.

To maximize unit economics:

Leverage Prefix Caching: Structure your system prompts and tool definitions at the beginning of the context payload to guarantee cache hits across multi-turn interactions.
Implement Batch Processing: For non-realtime operations such as data evaluation, document parsing, and backfill tasks, use asynchronous Batch APIs to immediately slash input and output bills in half.
Monitor Token Bloat: Uncontrolled conversational turns degrade both latency and token economics. Prune older history or summarize intermediate thoughts before context exceeds 32k tokens.

Frequently Asked Questions

What is the token pricing for Mistral Small 3 in 2026?

Mistral Small 3 is billed at $0.20 per million input tokens and $0.60 per million output tokens under standard API tiers.

What is the context window capacity of Mistral Small 3?

Mistral Small 3 supports up to 32,768 tokens, allowing long-document ingestion and extensive conversational history.

Does Mistral Small 3 support prompt caching or batch discounts?

Yes, prompt caching reduces recurring prefix token costs to roughly $0.10 per million tokens for cached blocks.

Related Frontier Showdowns & Hubs

Mistral Small 3: 24B Architecture & Sub-$0.10 High-Efficiency Inference TCO

Mistral's Mistral Small 3 (24B parameters) is an enterprise workhorse model designed for high-throughput classification, structured entity extraction, and conversational customer support. Priced at just $0.10 per 1M input tokens and $0.30 per 1M output tokens with a 32,768-token context window, it delivers performance matching Llama-3.3-70B while requiring less than half the GPU hardware footprint for self-hosting.

Model Comparison	Input / 1M Tokens	Output / 1M Tokens	Parameters	Monthly Cost (1B Tokens)
Mistral Small 3	$0.10 / 1M	$0.30 / 1M	24 Billion	$200.00 / month (50:50 ratio)
Llama-3.3-70B (Together Turbo)	$0.88 / 1M	$0.88 / 1M	70 Billion	$880.00 / month (4.4× higher)
Claude 3.5 Haiku	$0.80 / 1M	$4.00 / 1M	Proprietary (~20B)	$2,400.00 / month (12× higher)

Production Pipeline: High-Throughput JSON Classification Client

Classifying user support tickets into structured routing categories with Mistral Small 3:

from mistralai import Mistral

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

def classify_ticket(ticket_body: str):
    response = client.chat.complete(
        model="mistral-small-latest",
        messages=[
            {"role": "system", "content": "Classify the ticket into priority (P1-P4) and category (billing, technical, auth). Output strict JSON."},
            {"role": "user", "content": ticket_body}
        ],
        response_format={"type": "json_object"},
        temperature=0.0
    )
    return response.choices[0].message.content

Production Best Practices & Self-Hosting Guidelines

Single-GPU Serving: When quantized to FP8, Mistral Small 3 consumes roughly 25GB of VRAM, fitting onto a single NVIDIA L40S or RTX 3090/4090 with zero tensor-parallelism network communication overhead.
High Batching Efficiency: Due to its compact 24B parameter size, vLLM can sustain concurrent batch sizes exceeding 64 requests per GPU, delivering over 1,200 aggregate tokens per second.
Multi-Cloud Availability: Available on Mistral La Plateforme, Azure AI Studio, and AWS Bedrock with unified OpenAI-compatible endpoints.

24B Parameter Production Sizing & High-Throughput Batch Inference

Mistral Small 3 delivers 24 billion parameters of dense transformer architecture engineered to sit at the sweet spot between lightweight 8B models and massive 70B clusters. In FP16 precision, Mistral Small 3 requires 48 GB of memory, making it an ideal fit for single-GPU deployments on enterprise NVIDIA L40S or RTX 6000 Ada accelerators. When served via vLLM with FP8 weight quantization, memory consumption drops below 26 GB, enabling full 32k context processing with large concurrent batch sizes on a single affordable GPU node.

Mistral Small 3 Serving Configurations & Cost Analysis

Deployment Configuration	VRAM Footprint	Context Window	Sustained Tokens/Sec	Compute Cost / 1M Tokens
Managed Mistral AI API	Serverless	32,768 tokens	85 tokens/s	$0.20 input / $0.60 output
Self-Hosted 1x L40S (FP8)	26.2 GB	32,768 tokens	115 tokens/s	$0.065 / 1M (at 80% util)
Self-Hosted 2x L4 (INT4)	14.8 GB	16,384 tokens	72 tokens/s	$0.048 / 1M (at 80% util)

Production Function Calling Integration in Python

Invoke Mistral Small 3 with structured tool declarations for reliable autonomous routing:

from mistralai import Mistral

client = Mistral(api_key=os.environ["MISTRAL_API_KEY"])

tools = [{
    "type": "function",
    "function": {
        "name": "lookup_stock_ticker",
        "description": "Fetch real-time financial metrics for an equity ticker",
        "parameters": {
            "type": "object",
            "properties": {"ticker": {"type": "string"}},
            "required": ["ticker"]
        }
    }
}]

response = client.chat.complete(
    model="mistral-small-latest",
    messages=[{"role": "user", "content": "What is the valuation of NVDA?"}],
    tools=tools
)
print(response.choices[0].message.tool_calls)