TechPrufer
Get started
Mistral AI
Mistral AI logoMistral AI · Feature trail

Model Customization and Agents

11 posts · Dec 2023 – Mar 2026

What this trail covers

Mistral AI Forge enables enterprises to build frontier-grade AI models grounded in proprietary knowledge through pre-training, post-training, and reinforcement learning. This allows for custom models that understand enterprise-specific terminology, workflows, and policies, enhancing the reliability and accuracy of enterprise agents. The platform integrates with Mistral AI Studio for building, evaluating, and running AI in production, leveraging Temporal for durable and reproducible agent execution. Mistral AI is also introducing model customization on la Plateforme, offering open-source fine-tuning SDKs (mistral-finetune) using LoRA adapters, serverless fine-tuning services on la Plateforme, and custom training services for select customers. These services aim to reduce the cost and expertise required for customizing generative AI models at scale, enabling users to adapt Mistral AI models to their specific needs.

All Mistral AI trails
  1. Introducing Forge | Mistral AIThis post introduces Forge, a new system for enterprises to build custom AI models grounded in proprietary knowledge. It details the training methodologies supported (pre-training, post-training, reinforcement learning) and their application to enterprise-specific data like documentation, codebases, and operational records. The post emphasizes how custom models enhance enterprise agent reliability by providing deeper domain understanding, leading to more precise tool selection and reliable multi-step workflows. It also highlights support for dense and MoE architectures, multimodal inputs, and a continuous improvement lifecycle through RL and evaluation. The agent-first design philosophy is also mentioned, with Forge being usable by agents for hyperparameter tuning and synthetic data generation.
  2. Leanstral: Open-Source foundation for trustworthy vibe-coding | Mistral AIIntroduces Leanstral, an open-source code agent specifically for Lean 4. Details its sparse architecture (6B active parameters), evaluation methodology using the FLTEval suite, and comparative performance against other coding agents and open-source models. Provides case studies demonstrating its ability to answer Stack Exchange questions about Lean 4 changes and translate/reason about programs from Rocq. Highlights its availability through Mistral Vibe, a free API, and as downloadable weights.
  3. Rails testing on autopilot: Building an agent that writes what developers won't | Mistral AIThis post details the technical implementation of an autonomous agent for generating and improving RSpec tests in Rails applications. It describes the agent's architecture, which leverages Mistral's Vibe platform, including the use of `AGENTS.md` for step-by-step execution plans and specialized `SKILLS` files for different Rails file types (models, controllers, etc.). The post also elaborates on the integration of custom tools like RuboCop for linting and SimpleCov for test validation and code coverage, including stripped-down code examples for tool implementation. It further discusses metrics for measuring test quality, including tool-based metrics and the concept of 'LLM-as-a-judge' for qualitative assessment.
  4. Unlocking the potential of vision language models on satellite imagery through fine-tuning | Mistral AIThis post details the application of LoRA fine-tuning to Pixtral-12B for satellite imagery classification using the Aerial Image Dataset (AID). It demonstrates how LoRA can efficiently adapt model weights to specific tasks and domains, leading to significant improvements in classification accuracy (from 0.56 to 0.91) and a reduction in hallucinations. The post also highlights the ease of fine-tuning via Mistral's API and LaPlateforme UI, and provides guidance on hyperparameter selection. It showcases the effectiveness of domain-specific adaptation for specialized visual domains like satellite imagery.
  5. Upgrading agentic coding capabilities with the new Devstral models | Mistral AIIntroduces Devstral Small 1.1 and Devstral Medium, enhancing agentic coding capabilities. Devstral Small 1.1, an upgrade to the previous version, is released under Apache 2.0 and achieves 53.6% on SWE-Bench Verified, showing improved generalization and support for Mistral function calling and XML formats. Devstral Medium, available via API, scores 61.6% on SWE-Bench Verified and offers competitive cost/performance, with options for on-premise deployment and custom fine-tuning.
  7. Mistral Saba | Mistral AIIntroduces Mistral Saba, a 24B parameter model specialized for Middle Eastern and South Asian languages (Arabic, South Indian languages like Tamil). Highlights its performance advantage over larger models, speed, cost-effectiveness, and local deployability on single-GPU systems. Mentions its use as a base for further regional adaptations and its availability via API and for private customer deployments.
  8. Mistral Small 3 | Mistral AIIntroduces Mistral Small 3, a 24B-parameter model optimized for latency and local deployment. Highlights its performance benchmarks against larger models and proprietary alternatives, and details its suitability for conversational AI, function calling, fine-tuning, and local inference. The post also reiterates Mistral AI's commitment to Apache 2.0 licensing for general-purpose models and mentions upcoming releases with boosted reasoning capabilities.
  9. Codestral 25.01 | Mistral AICodestral 25.01 introduces a more efficient architecture and an improved tokenizer, leading to a 2x speed improvement in code generation and completion. It achieves SOTA performance for FIM use cases and offers a 256k context length. Benchmarks show significant improvements over previous versions and competitors across various coding tasks and languages.
  10. Build, tweak, repeat | Mistral AIThis post announces the ability to customize Mistral AI's flagship and specialist models (including Mistral Large 2 and Codestral) on La Plateforme using base prompts, few-shot prompting, or fine-tuning with custom datasets. It also introduces an early alpha release of Agents, which wraps models with additional context and instructions for Le Chat or API to create custom behaviors and workflows. Additionally, the `mistralai` client SDK has been released as version 1.0 for Python and Typescript.
  11. My Tailor is Mistral | Mistral AIThis post introduces three new entry points for specializing Mistral AI models: an open-source fine-tuning SDK (mistral-finetune) built on LoRA for developers to fine-tune on their own infrastructure, serverless fine-tuning services on la Plateforme leveraging Mistral's proprietary techniques (also using LoRA adapters), and custom training services for select customers involving continuous pretraining. The post highlights the efficiency and cost-effectiveness of LoRA adapters for fine-tuning Mistral 7B and Mistral Small models, showing comparable performance to full fine-tuning with reduced costs for training and serving.
  12. La Plateforme | Mistral AIThis post announces the beta launch of 'La Plateforme', Mistral AI's first platform services, providing API endpoints for text generation and embeddings. It details the models served by each endpoint (mistral-tiny, mistral-small, mistral-medium, mistral-embed), their performance benchmarks (MT-Bench, MTEB), and the underlying alignment techniques used. It also mentions API specifications, client libraries, and the ramp-up process towards general availability, acknowledging NVIDIA's contribution to optimizing sparse mixture of experts models.