Skip to content

Magistral Small 2509

Magistral Small 2509 is a 24B open-source reasoning model under Apache 2.0 with 70.7% on AIME 2024. It's Mistral AI's compact reasoning option in the Magistral family.

ReasoningVision (Image)
index.ts
import { streamText } from 'ai'
const result = streamText({
model: 'mistral/magistral-small',
prompt: 'Why is the sky blue?'
})

Playground

Try out Magistral Small 2509 by Mistral AI. Usage is billed to your team at API rates. Free users (those who haven't made a payment) get $5 of credits every 30 days.

Providers

Route requests across multiple providers. Copy a provider slug to set your preference. Visit the docs for more info. Using a provider means you agree to their terms, listed under Legal.

Provider
Context
Latency
Throughput
Input
Output
Cache
Web Search
Per Query
Capabilities
ZDR
No Training
Release Date
Mistral AI
Legal:Terms
Privacy
128K
0.2s
112tps
$0.50/M$1.50/M
03/17/2025
Throughput

P50 throughput on live AI Gateway traffic, in tokens per second (TPS). Visit the docs for more info.

Latency

P50 time to first token (TTFT) on live AI Gateway traffic, in milliseconds. View the docs for more info.

Uptime

Direct request success rate on AI Gateway and per-provider. Visit the docs for more info.

More models by Mistral AI

Model
Context
Latency
Throughput
Input
Output
Cache
Web Search
Per Query
Capabilities
Providers
ZDR
No Training
Release Date
256K
0.3s
76tps
$0.40/M$2.00/M
mistral logo
12/09/2025
256K
0.4s
70tps
$0.50/M$1.50/M
mistral logo
12/02/2025
256K
0.2s
103tps
$0.20/M$0.20/M
mistral logo
12/01/2025
128K
0.2s
199tps
$0.10/M$0.10/M
mistral logo
10/01/2024
32K
0.3s
174tps
$0.10/M$0.30/M
mistral logo
09/01/2024
$0.10/M
mistral logo
12/11/2023

About Magistral Small 2509

Magistral Small 2509 was announced alongside Magistral Medium on March 17, 2025 as the open-source counterpart in Mistral AI's first reasoning model release. At 24B parameters, it scores 70.7% on AIME 2024 and 83.3% with majority voting.

Like Magistral Medium, Magistral Small 2509 produces transparent chain-of-thought reasoning and covers the same eight languages: English, French, Spanish, German, Italian, Arabic, Russian, and Simplified Chinese. It applies reasoning depth to business strategy, regulated industries, software engineering, and creative applications, rather than functioning as a smaller instruct model.

The Apache 2.0 license means weights are available on Hugging Face for commercial use.

What To Consider When Choosing a Provider

  • Configuration: Magistral Small 2509's Apache 2.0 license makes it the only reasoning model in the Magistral family that can be freely incorporated into commercial products without a separate licensing agreement.
  • Zero Data Retention: AI Gateway supports Zero Data Retention for this model via direct gateway requests (BYOK is not included). To configure this, check the documentation.
  • Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.

When to Use Magistral Small 2509

Best For

  • Apache 2.0 commercial products: Need this license for reasoning model integration
  • Cost-sensitive reasoning workloads: Applications where the 70.7% AIME 2024 score is sufficient
  • Research and educational use: Cases requiring transparent, auditable reasoning

Consider Alternatives When

  • Highest reasoning accuracy: You need the top option (consider Magistral Medium at 73.6%)
  • No structured reasoning needed: The task does not require multi-step reasoning at all

Conclusion

Magistral Small 2509 reaches 70.7% on AIME 2024 with transparent chain-of-thought and Apache 2.0 licensing. Access it through AI Gateway for reasoning workloads without proprietary model lock-in.

Frequently Asked Questions

  • What are Magistral Small 2509's AIME benchmark scores?

    70.7% on AIME 2024 and 83.3% with majority voting.

  • What is the parameter count and license?

    24B parameters under Apache 2.0, available for download from Hugging Face.

  • Does Magistral Small 2509 produce chain-of-thought reasoning?

    Yes. Like Magistral Medium, Magistral Small 2509 exposes step-by-step reasoning traces for each response.

  • Which languages does Magistral Small 2509 reason in?

    English, French, Spanish, German, Italian, Arabic, Russian, and Simplified Chinese, the same set as Magistral Medium.

  • How does Magistral Small 2509 compare to Magistral Medium?

    Magistral Small 2509 is 24B parameters and open-source. Magistral Medium is the larger, API-first release. On AIME 2024, Small is 70.7% single-sample versus Medium at 73.6%; the gap is most visible on heavy math, less on typical business prompts.

  • Is Magistral Small 2509 available via API?

    Yes. You can access Magistral Small 2509 through AI Gateway.