NOUMENOpen app
NOUMEN

Mistral · Open weights

Mistral Large 4

Added Oct 6, 2026Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Mistral Large 4 is an open-weight multimodal mixture-of-experts model with 1.05T total and 49B active parameters, built for reasoning, coding, tool use, and structured output. It supports a 512K-token context window.

  • Reasoning
  • Tools
  • Vision
  • Structured output
Input
Text, Images
Output
Text
Reasoning effort
none · high
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.68
Output
$2.09
Cache read
$0.07

Reference prices as published by the model provider. API example

Specifications

Context window
524.3K
Max output
—

Side by side

Mistral Large 4 compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Mistral Large 4MistralThis modelKimi K2.7 CodeMoonshotRemove
ProviderMistralMoonshot
Discovery tierOpen weightsOpen weights
AddedOct 6, 2026Jun 12, 2026
Limits
Context window524.29KBest262.14K
Max output—65.54K
Average output—1.8K tokens
Pricing, USD per 1M tokens
Input$0.68$0.67Best
Output$2.09Best$3.35
Cache read$0.07Best$0.15
Performance
Throughput—63.9 tok/s
Time to first token—4.6 s
Request success (24h)68.9%99.6%Best
Benchmark indexes
Intelligence Index—25.8
Coding Index—60.8
Capabilities
ReasoningYesYes
ToolsYesYes
VisionYesYes
Structured outputYesNo
Parallel toolsNoYes
Reasoning effortnone, high—

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Mistral Large 4
  • Kimi K2.7 Code

Intelligence Index

  • Mistral Large 4—
  • Kimi K2.7 Code25.8
3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Mistral Large 4: —
    • Kimi K2.7 Code: 60.8

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Mistral Large 4: —
    • Kimi K2.7 Code: 24.5%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Mistral Large 4: —
    • Kimi K2.7 Code: 4.7%

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Mistral Large 4: —
    • Kimi K2.7 Code: 858 Elo

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Mistral Large 4: —
    • Kimi K2.7 Code: 1,038 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Mistral Large 4: —
    • Kimi K2.7 Code: 11.2%

  • AA-LCRv1.1

    Reasoning across long documents

    • Mistral Large 4: —
    • Kimi K2.7 Code: 79.3%

  • MLCR-AA

    Long-context reasoning over medical records

    • Mistral Large 4: —
    • Kimi K2.7 Code: 12.2%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Mistral Large 4: —
    • Kimi K2.7 Code: 35.0%

  • GPQA Diamond

    Graduate-level science questions

    • Mistral Large 4: —
    • Kimi K2.7 Code: 89.6%

  • IFBench

    Following precise output instructions

    • Mistral Large 4: —
    • Kimi K2.7 Code: 63.1%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Mistral Large 4: —
    • Kimi K2.7 Code: 1.0%

  • SciCode

    Scientific computing in Python

    • Mistral Large 4: —
    • Kimi K2.7 Code: 47.8%

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Mistral Large 4: —
    • Kimi K2.7 Code: 44.7%

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Mistral Large 4: —
    • Kimi K2.7 Code: 90.1%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Mistral Large 4: —
    • Kimi K2.7 Code: 79.3%

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Prompt cache hits
57.2%
About $0.33 per 1M input tokens with caching
Request success (24h)
68.8%
Failing

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 5.1%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mistralai/mistral-large-4",
    "reasoning_effort": "none",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="mistralai/mistral-large-4",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Mistral

Other Mistral models, starting with the same family.