Last verified
RETIRED JUL 2026128K CONTEXTPREMIER TIERNO CACHE SKU

Magistral Medium API Pricing

Magistral Medium is retired. Mistral publishes a retirement date of July 31, 2026 for MAGISTRAL-MEDIUM-2509 and names Mistral Medium 3.5 as the alternative; the separate reasoning line has been folded into the generalist models. This archive preserves the last published card: $2/M input and $5/M output, with no separate prompt-cache SKU.

Input - per 1M tokens
$2.00/M
Source Mistral flat
Output - per 1M tokens
$5.00/M
Context 128K flat
Cache N/A
$2.00/M
Cache no separate SKU not listed
Effective - agentic blend
$2.24/M
92/8 split - no cache
§ 01 / TERMINAL

Magistral Medium API cost calculator.

Live calculator pre-loaded with Magistral Medium rates. Tweak spend, output mix, or cache assumptions to compare this row against current Mistral replacements.

$ /mo
Workload split
Tokens you can process
—
Words equivalent (English)
—
Effective rate
—
Open full calculator (all models · share URL · CSV) →
§ 02 / SCENARIOS

Magistral Medium cost on real workloads.

CHATBOT

Support assistant

$0.0065/turn
2,000 in - 500 out~15,384 units/$100
CODING AGENT

Repo task iteration

$0.125/task
50,000 in - 5,000 out~800 units/$100
LONG DOC

Contract review

$0.370/doc
180,000 in - 2,000 out~270 units/$100
RAG

Knowledge base answer

$0.021/query
8,000 in - 1,000 out~4,761 units/$100
§ 03 / TOKENIZER

Magistral Medium token counter and cost.

What's a token →
Calibrated · measured on the vendor's tokenizer · 2026-06-10 Auto-counts as you type

Counts use a chars-per-token calibration measured on the vendor's own published tokenizer (mistralai/Mistral-Nemo-Instruct-2407, 2026-06-10). English prose is typically within a few percent; code and non-Latin scripts tokenize heavier. For billing-exact counts use the vendor's count-tokens API.

Characters 408
Words 65
Tokens (estimated) 80 tokens
Cost as input · uncached $0.00016 USD
Cost as output · uncached $0.0004 USD
§ 04 / SHELF

Magistral Medium vs similar models.

All models →
Model Input /M Output /M Effective blended Context Best for
Magistral Medium Current $2.00 $5.00 $2.24 agentic 92/8 documented elsewhere — not on pricing page Transparent reasoning
Mistral Large 3 $0.50 cache $0.05 $1.50 $0.241 −9.3× documented elsewhere — not on pricing page Frontier multimodal work
Mistral Medium 3.5 $1.50 cache $0.15 $7.50 $0.962 −2.3× documented elsewhere — not on pricing page General Mistral workloads
Mistral Small 4 $0.15 cache $0.015 $0.60 $0.0842 −27× 256K Low-cost chat and support
Codestral $0.30 cache $0.03 $0.90 $0.144 −16× documented elsewhere — not on pricing page Code generation and agents
Devstral 2 $0.40 $2.00 $0.528 retired documented elsewhere — not on pricing page Code generation and agents
DeepSeek V4.1 Flash $0.30 cache $0.006 $1.20 $0.15 −15× 1M Budget coding and bulk reasoning
§ 07 / BENCHMARKS

Magistral Medium benchmarks

All benchmarks →

Magistral Medium has published scores on 2 benchmarks we track; its best placing is #91 of 99 on ARC-AGI-1. All of these scores come from independent leaderboards, not from the lab.

Benchmark Score Rank Leader Source
ARC-AGI-1 Reasoning 5.9% ARC Prize verified, best result #91 of 99 6 tied · 98.5% Official leaderboard ARC Prize Foundation
ARC-AGI-2 Reasoning 0% ARC Prize verified, best result #97 of 105 GPT-6 Astra · 95% Official leaderboard ARC Prize Foundation

Magistral Medium API pricing FAQ.

Practical pricing questions, with archived prices separated from current availability.

Q · 01 What is Magistral Medium priced at? +
Magistral Medium is shown here at $2/M input and $5/M output. The current official model card shows the same headline rate.
Q · 02 Is Magistral Medium still available? +
Magistral Medium is listed as an active Mistral model today. The page keeps the live price and calculator assumptions together for invoice checks.
Q · 03 Does Mistral list prompt caching for this model? +
No separate prompt-cache SKU is published for this row. The calculator therefore treats cache-hit input as the base input rate, $2/M, so the effective blend does not assume a hidden discount.
Q · 04 What does the effective price mean? +
The effective tile uses the site-wide agentic blend: 92% input tokens, 8% output tokens, and no cache discount unless a vendor lists one. For Magistral Medium, that gives $2.24/M.
Q · 05 Where does the archive price come from? +
For active rows, the price comes from Mistral's current pricing or model card. For retired rows, AI//COST preserves the verified snapshot price and uses Mistral's current docs to verify deprecation, retirement, and replacement status.
Q · 06 Is there a batch discount? +
Mistral exposes batching in the API docs, but this archive row does not publish a separate per-model batch rate. The calculator keeps the baseline token prices visible and avoids inventing a discount.
Y
Reviewed by Yaroslav Vikhariev Founder · AI//COST · Prices checked by hand against mistral.ai · Last verified Aug 23, 2026
Methodology Report a correction More by Y.V.