Ministral 3 14B (v25.12, Base/Instruct/Reasoning)
Ministral 3 14B, developed by MistralAI, is a multimodal foundation model with advanced reasoning capabilities. It is equipped with a long context window of 256K and is released under the Apache 2.0 license.
Parameters
140.0B
Context Window
256K
License
Apache 2.0
Release Date
2025-12-02
API Pricing
API pricing for this model is not yet available
Strengths
- ・Possesses advanced reasoning capabilities
- ・256K long context processing
- ・Open license
Weaknesses
- ・Requires computational resources of 140B scale
- ・Load for multimodal processing
- ・Possibility of increased operational costs
Use Cases
- ・Complex logical reasoning tasks
- ・Analysis and summarization of long documents
- ・Multimodal data processing
Deep Analysis
Architecture
Dense Transformer (14B)
Largest in Ministral 3 family
Context Window
262K tokens
API Pricing (Input/Output)
$0.20 / $0.20 per 1M tokens
Blended $0.40/1M
Release Date
December 2, 2025
License
Open weights (v25.12)
Output Speed
78-181 t/s
Varies by provider
Strengths
- ・Frontier capabilities at small model size
- ・Very affordable pricing ($0.20/1M)
- ・262K context window
- ・Open weights
- ・Good balance of speed and performance
Weaknesses
- ・Smaller than Mistral Large models
- ・Limited multimodal support
- ・Reasoning depth may not match larger models
Competitor Comparison
| Model | Arena | SWE | GPQA | Price |
|---|---|---|---|---|
| Ministral 8B | - | - | - | $0.10/$0.10 |
| Mistral Small 4 | - | - | - | Higher |
| Llama 3.3 70B | - | - | - | Higher |
Ministral 3 14B is the largest model in Mistral's Ministral 3 family, offering frontier capabilities in a compact 14B parameter package. Released December 2025 with 262K context and very competitive pricing.
Analysis generated: 2026-05-24