Mistral Medium 3.5
Mistral's 128B dense merged model unifying instruction-following, reasoning, and coding with configurable reasoning effort
Mistral Medium 3.5
Mistral AI • April 2026
Training Data
Not disclosed
Mistral Medium 3.5
April 2026
Parameters
128 billion (dense)
Training Method
Merged dense model with configurable reasoning-effort
Context Window
256,000 tokens
Knowledge Cutoff
Not disclosed
Key Features
Configurable Reasoning Effort • Native Vision Encoder • 256K Context • Modified MIT License
Capabilities
Coding: Outstanding
Reasoning: Excellent
Vision: Excellent
What's New in This Version
Replaces Devstral 2 and Magistral by merging instruction, reasoning, and coding into one 128B dense model; SWE-Bench Verified 77.6%, τ³-Telecom 91.4; powers Mistral Vibe remote coding agents
Mistral's 128B dense merged model unifying instruction-following, reasoning, and coding with configurable reasoning effort
What's New in This Version
Replaces Devstral 2 and Magistral by merging instruction, reasoning, and coding into one 128B dense model; SWE-Bench Verified 77.6%, τ³-Telecom 91.4; powers Mistral Vibe remote coding agents
Technical Specifications
Key Features
Capabilities
Other Mistral AI Models
Explore more models from Mistral AI
Mistral Small 4
Mistral's unified open-weight MoE model combining reasoning, multimodal, and coding capabilities under Apache 2.0 with only 6.5B active parameters
Mistral Large 3
Mistral's state-of-the-art open-weight frontier model with multimodal and multilingual capabilities under Apache 2.0
Ministral 3 14B
Mistral's high-performance dense model in the new Ministral 3 family
