When was Mistral Small 4 released?

Mistral released Mistral Small 4 on 16 Mar 2026. It is available through the qynio gateway.

Does Mistral Small 4 support zero data retention?

Yes. Mistral Small 4 is available with zero data retention by default on Regolo through the qynio gateway.

Where is Mistral Small 4 hosted?

Mistral Small 4 runs on crypto rails routes through the qynio gateway, with contractually enforced residency on supported providers.

Is Mistral Small 4 trained on prompts and outputs?

No. Every hosting route for Mistral Small 4 is configured so customer prompts and outputs are not used to train models by default.

Is a no-KYC deposit available for Mistral Small 4?

Yes. A no-KYC Data Processing Agreement is available on at least one route serving Mistral Small 4. Contact hello@qyn.io to receive it.

Mistral Small 4

by Mistral

Released 104 days after Mistral Large 3 · All Mistral releases

Mistral Small 4, released March 16 2026, is a Mixture-of-Experts model that consolidates the capabilities of Magistral (reasoning), Devstral (coding), and Mistral Small (instruct) into one versatile system. It has roughly 119B total parameters with about 6B active per token, so it stays efficient at inference while eliminating the need to choose between a fast chat model, a reasoning engine, and a multimodal assistant. With a 256K token context window and Apache 2.0 licensing, Small 4 accepts both text and image inputs for general chat, agentic tasks, and complex reasoning. A reasoning effort setting lets you trade response speed against depth. With reasoning enabled, Small 4 matches or surpasses GPT-OSS 120B on several reasoning benchmarks while producing significantly shorter outputs. The model is available across common inference stacks including NVIDIA NIM, vLLM, SGLang, llama.cpp, and Hugging Face Transformers, supporting both cloud and on-premise deployment. Small 4 is built for agentic AI, reasoning workflows, and code-intensive applications where a single adaptable model reduces operational complexity.

Open in playground Try in chat Compare

Key info

Input

Output

Features

Context window

256K

Max output

Input price

$0.15 /1M

Output price

$0.60 /1M

Released

16 Mar 2026

onchain settlement available
Zero data retention on pay-as-you-go
No training by default
no-KYC deposit available

Available routes

Mistral Small 4 runs on 2 different routes through the qynio gateway. Compare residency, zero retention, and training posture at a glance — full data-handling detail per route below.

Provider	Region	Zero data retention	Training	Input	Output
Mistral	EU	Not by default	No	$0.15	$0.60
Regolo	EU	Zero data retention	No	$0.58	$2.44

Uptime and availability

One of Mistral Small 4's 2 routes is on the monitoring board, so the figure below is that provider's availability rather than the model's.

100%route uptime, last 30 days

Its other routes aren't measured here yet.

Name a second model in the same request and the gateway tries it on retriable errors, so a busy hour never has to reach your users. .

Measured over the last 30 days from each provider's official status feed via StatusGator. Refreshed hourly. 1 further route is not on the board yet. See uptime for every provider qynio monitors.

Data handling per route

Each route hosting Mistral Small 4 has its own privacy posture, residency, and no-KYC terms. Postures are maintained by qynio with a last-verification timestamp.

Mistral — France🇫🇷

Zero data retention is not on by default — abuse monitoring, 30-day retention. No training on customer data. EU; deposit available.

Zero data retention: Not by default — abuse monitoring, 30-day retention. Derived from the logging and moderation facts.
Training: No training on customer data.
Logging: Abuse monitoring (30-day retention)
Moderation: Not established
Caching: Not established
Subprocessor access: Not established
no-KYC deposit: deposit available
Transfer mechanism: Not applicable — data stays in EU

Regolo — Italy🇮🇹

Zero data retention is on by default on Pay-as-you-go — no action required. No training on customer data. EU; deposit available.

Zero data retention: On by default on Pay-as-you-go. Derived from the logging and moderation facts.
Training: No training on customer data.
Logging: None
Moderation: No hold on content
Caching: Not established
Subprocessor access: No subprocessor reads content
no-KYC deposit: deposit available
Transfer mechanism: Not applicable — data stays in EU

Benchmarks

Independent benchmark scores — composite indices for reasoning, coding, and math, plus individual eval scores where available.

Global rank#285 of 630 LLMs

TierEfficient

Output speed164 tok/s

First token0.63s

Intelligence Index19.7

Coding Index26.6

Reasoning & knowledge

GPQA Diamond

77%

Humanity's Last Exam

10%

Long-context reasoning

47%

Coding

SciCode

38%

Agentic & tool use

Terminal-Bench Hard

17%

τ²-Bench Telecom

41%

Math & instruction following

IFBench

48%

See full leaderboard →Benchmarks via Artificial Analysis · View on AA

Get started

Call Mistral Small 4 through the qynio gateway with one API key. Let your coding agent set it up, or call it directly — qynio is drop-in compatible with the OpenAI, Anthropic, and Google AI SDKs.

Set it up with your agent

Copy this and paste it into a coding agent like Claude Code, Cursor or Codex and it'll wire up qynio for you.

Or call it directly

1import OpenAI from "openai";
2
3const client = new OpenAI({
4  apiKey: process.env.QYN_API_KEY,
5  baseURL: "https://api.qyn.io/v3/compat",
6});
7
8const completion = await client.chat.completions.create({
9  model: "mistral/mistral-small-2603",
10  messages: [{ role: "user", content: "Hello" }],
11});
12console.log(completion.choices[0].message.content);