Mistral Large: Complete Specifications, Pricing, API Access & Use Cases (2026)

What Is Mistral Large?

Mistral Large is Mistral AI’s flagship general-purpose model line; its current verified API variant, Mistral Large 3, was released on December 2, 2025, featuring a 256K-token context window and multimodal text-and-image understanding, with API pricing of \$0.50 per 1M input tokens and \$1.50 per 1M output tokens as of June 2026.

The name "Mistral Large" originally referred to Mistral AI’s February 26, 2024 flagship text-generation model, which had a 32K-token context window and was available through Mistral’s platform and Azure. Mistral Large 2 followed in July 2024 with a 128K context window, broader multilingual support, code support, and the API name mistral-large-2407; Mistral’s current model overview now lists Mistral Large 2.0 as a legacy/deprecated model with Mistral Large 3 as its alternative.

Users usually search for Mistral Large to compare its long-context capability, open-weight availability, multimodal support, pricing, and API access against general-purpose models such as GPT-4o and Claude Sonnet-class models.

What Are Mistral Large’s Key Specifications and Pricing?

The table below uses the current verified Mistral Large 3 API variant where model-specific values are available. Pricing should be checked in the provider dashboard before production deployment because Mistral’s model card and launch card show \$0.50/\$1.50 for Mistral Large 3, while the general pricing FAQ still includes a generic "Mistral Large" example at \$2/\$6.

What Can Mistral Large Do That Makes It Useful in Production?

Mistral Large is relevant for production workflows that need a mix of long context, multilingual reasoning, multimodal document understanding, and API-based automation.

First, it can process long prompts and documents up to a 256K-token context limit, making it useful for contract review, long policy analysis, technical documentation, and multi-file knowledge workflows. Requests exceeding the model’s context window return an error, and token counts include both input and output tokens, so teams still need prompt budgeting.

Second, Mistral Large 3 supports image understanding through Mistral’s vision-capable Chat Completions flow, which can be useful for chart interpretation, screenshot analysis, document-image triage, and multimodal support tickets. Mistral’s vision docs list Mistral Large 3 via mistral-large-2512 among recommended vision-capable models.

Third, it supports function calling and structured outputs, which matters when the model must call internal tools, generate JSON, classify records, or return schema-aligned data for downstream systems. Function calling helps the model decide when to request external data, but developers remain responsible for executing tools, validating outputs, and handling errors.

Fourth, its open-weight Apache 2.0 release can matter for teams comparing hosted API access with self-hosted or private deployment options. Self-hosting still requires substantial infrastructure, inference engineering, monitoring, and security review.

What Are Mistral Large’s Supported Modalities?

Where Does Mistral Large Fall Short?

Mistral Large has several important limitations.

• The first is source and naming complexity. "Mistral Large" can refer to the original 2024 model, Mistral Large 2, the mistral-large-latest alias, or the current Mistral Large 3 model card. Developers should pin a model ID, such as mistral-large-2512, when reproducibility matters.
• The second is pricing ambiguity. The Mistral Large 3 model card and launch card show \$0.50 per 1M input tokens and \$1.50 per 1M output tokens, while Mistral’s general pricing FAQ still displays a \$2/\$6 example for "Mistral Large." Treat the dashboard or model-specific pricing as the final billing source before deployment.
• The third is general AI reliability. This is a general AI limitation and is not model-specific unless stated by the provider: LLMs can hallucinate, miss context, produce insecure code, or misread ambiguous prompts. High-stakes legal, medical, financial, or safety decisions require expert review.
• The fourth is platform constraint. Rate limits vary by organization, model, and subscription tier, and Mistral documents that requests per second, tokens per minute, and monthly token caps can apply independently.

What Is Mistral Large Best Used For?

Teams evaluating broad multimodal APIs often compare Mistral Large with GPT-4o specs and API access for OpenAI-centered workflows and Claude 3.5 Sonnet specs and API access for long-context assistant workflows. For open-weight alternatives, Llama 3.1 405B Instruct specifications provide a useful comparison point.

How Does Mistral Large Compare to GPT-4o and Claude 3.5 Sonnet?

This comparison is scenario-qualified, not a ranking. GPT-4o remains a useful multimodal reference point, Claude 3.5 Sonnet remains useful for long-context assistant comparisons, and Mistral Large 3 is especially relevant when users want a current open-weight model with long context and provider API access.

How Do I Access Mistral Large?

Developers can access the current Mistral Large API through Mistral’s Chat Completion endpoint using the model alias mistral-large-latest or the pinned model ID mistral-large-2512. Mistral’s API reference verifies POST /v1/chat/completions, bearer authentication, the messages request field, model selection through the model parameter, and response objects containing choices, model, and usage.

Python Example

from mistralai.client import Mistral
import os
api_key = os.environ["MISTRAL_API_KEY"]
with Mistral(api_key=api_key) as client:
    response = client.chat.complete(
        model="mistral-large-latest",
        messages=[
            {
                "role": "user",
                "content": "Summarize the main deployment risk in one sentence."
            }
        ],
        stream=False,
        response_format={"type": "text"},
    )
print(response.choices[0].message.content)

curl Example

curl https://api.mistral.ai/v1/chat/completions \
  -X POST \
  -H "Authorization: Bearer $MISTRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mistral-large-latest",
    "messages": [
      {
        "role": "user",
        "content": "Summarize the main deployment risk in one sentence."
      }
    ],
    "response_format": {
      "type": "text"
    }
  }'

For repeated long prompts, Mistral supports prompt caching through prompt_cache_key; cached prompt tokens are billed at 10% of the standard input token price when cache hits occur. For asynchronous large workloads, Mistral’s Batch Processing supports Chat Completions and offers a 50% discount.

FAQs

What is the context window of Mistral Large?
The current verified Mistral Large 3 model card lists a 256K-token context window as of June 2026. Older Mistral Large versions had smaller limits, so production systems should pin the model ID.

How much does Mistral Large cost?
Mistral Large 3 model-card pricing is \$0.50 per 1M input tokens and \$1.50 per 1M output tokens as of June 2026. Cached input tokens are billed at 10% of standard input price when cache hits occur.

How do I use Mistral Large through the API?
Use Mistral’s Chat Completion endpoint, POST /v1/chat/completions, with bearer authentication, a model value such as mistral-large-latest, and a messages array.

What is Mistral Large best used for?
Mistral Large may fit long-document analysis, multilingual assistants, vision-assisted document workflows, tool-using applications, and teams evaluating open-weight deployment options. It should be tested against task-specific accuracy, latency, and cost requirements.