Mistral released a GPT-4-class model that costs 8× less than the competition.
What this means for you
Mistral Medium 3 achieves 90%+ of Claude Sonnet 3.7's benchmark scores at $0.40 input / $2.00 output per million tokens. The nearest comparable models (GPT-4o, Claude Sonnet) cost 3–8× more. Supports a 131K context window.
Bottom line impact
If you use GPT-4-class models at high volume and near-equivalent quality would suffice, switching to Mistral Medium 3 could cut your AI costs by 70–87%.
Should you care?
↗ Maybe
Yes if you use GPT-4-class models at volume and cost matters. Run a quality benchmark against your specific workload before switching — benchmarks don't always translate to your use case.
Test Mistral Medium 3 on your actual prompts via la Plateforme or AWS SageMaker. If quality meets your threshold, migrate high-volume, cost-sensitive workloads.
$870/mo per $1K spent on equivalent GPT-4 models
90%+ benchmark parity with models costing 8× more.
EU-based provider. Standard Mistral ToS applies.
Mistral Medium 3 announcement
https://mistral.ai/news/mistral-medium-3
Mistral API Pricing
https://mistral.ai/pricing
Mardii links only to official provider pages.
Free account. Follow the providers you care about. Unsubscribe anytime.