Mistral AI

Mistral AI: Devstral Small 2

deprecated

Devstral Small 2 is available through Everstack's OpenAI-compatible gateway with routing, fallback, retries, and observability.

labs-devstral-small-2512devstralreleased Dec 9, 2025
ProviderQuick start
Context
256K
tokens
Max output
256K
tokens
Input
-
per 1M
Output
-
per 1M

Activity

Daily period usage
Aggregated managed Everstack traffic. Buckets below 5 tenants or 100 requests are withheld.

Providers

ProviderContextInput /MOutput /M
Mistral AI
Mistral AI
256K--

Pricing

ItemPriceUnit
Input-per 1M tokens
Output-per 1M tokens

Performance

Reliability

Compare

Compare usage and performance
Place this model beside up to three other catalog models.

Capabilities

Capabilities
Modalities
Input
Output
Details
Model ID
labs-devstral-small-2512
Family
devstral
Status
deprecated
Knowledge
Dec 1, 2025
Released
Dec 9, 2025
Gateway Config
API
2024-01-01
Auth
MISTRAL_API_KEY
Discovery
static
Provider type
static

Quick Start

index.ts
import Everstack from "@everstack/node";const client = new Everstack({  apiKey: process.env.EVERSTACK_API_KEY!,  baseUrl: "https://{instance}.{region}.everstack.ai",});const completion = await client.chat.completions.create({  model: "@mistral/labs-devstral-small-2512",  messages: [{ role: "user", content: "Explain vector cache routing." }],});console.log(completion.choices[0]?.message?.content);

Endpoint

Base URL
https://api.mistral.ai/v1
Chat completions
https://api.mistral.ai/v1/chat/completions
Model
labs-devstral-small-2512
Auth env
MISTRAL_API_KEY

Devstral Small 2 FAQ

What is the context window of Devstral Small 2?

Devstral Small 2 accepts up to 256K tokens of context and returns up to 256K output tokens on Mistral AI.

What can Devstral Small 2 do?

Devstral Small 2 supports chat, function calling and image input. It accepts text and image input.

How do I call Devstral Small 2 through Everstack?

Point any OpenAI-compatible SDK at your Everstack gateway and set the model to @mistral/labs-devstral-small-2512. Everstack adds retries, provider fallback, caching, per-tenant budgets and OpenTelemetry traces for every request.