Compare AI models

Specifications
Parameters
—21B
Context window
128k128k
API pricing
Cheapest input
—$0.018AkashML
Cheapest output
—$0.09AkashML
—$8.10Darkbloom · fp8
Benchmarks
AIME 2024
91.82%92.1%
AIME 2025
83.48%91.7%
GPQA Diamond
76.26%71.5%
Overview
CompanyMistralOpenAI
Release dateSep 17 2025Aug 5 2025
AccessClosedOpen Weight
Model detailsView modelView model

Frequently asked questions

gpt-oss-20b leads Magistral Medium 1.2 on 2 of the 3 benchmarks they both report (AIME 2024, AIME 2025, GPQA Diamond). gpt-oss-20b shipped 43 days before Magistral Medium 1.2, so benchmark comparisons should account for the intervening progress.

Context windows are 128k (Magistral Medium 1.2) vs 128k (gpt-oss-20b). Magistral Medium 1.2 is closed, while gpt-oss-20b is open weight.

On AIME 2024, gpt-oss-20b leads at 92.1% vs Magistral Medium 1.2 at 91.82%. On AIME 2025, gpt-oss-20b leads at 91.7% vs Magistral Medium 1.2 at 83.48%. On GPQA Diamond, Magistral Medium 1.2 leads at 76.26% vs gpt-oss-20b at 71.5%.