# Post Trains

Before training our own models from scratch, Arcee AI was founded as a model post-training company. During this time, we focused on fine-tuning task specific models, taking open source models and enhancing them using training techniques pioneered by our research team. 

Find our previously released, and now retired, fine-tuned models below:

**General Purpose**

* [Arcee Blitz](#blitz)
* [Virtuoso Small](#virtuoso)
* [Virtuoso Medium](#virtuoso)
* [Virtuoso Large](#virtuoso)

**Reasoning**

* [Maestro](#maestro)

**Coding**

* [Coder Large](#coder)
* [Coder Small](#coder)

**Function Calling**

* [Caller](#caller)

{% tabs %}
{% tab title="Blitz" %}

### **Arcee Blitz**

* **Description:** Arcee-Blitz (24B) is a new Mistral-based 24B model distilled from DeepSeek, designed to be both **fast and efficient**. We view it as a practical “workhorse” model that can tackle a range of tasks without the overhead of larger architectures.
  * **#Parameters:** 24B
  * **Base Model:** Mistral-Small-24B-Instruct-2501
  * Open-source and available on Hugging Face under the Apache-2.0 license: [arcee-ai/Arcee-Blitz](https://huggingface.co/arcee-ai/Arcee-Blitz)
* **Top Use Cases:**
  * General-purpose task handling
  * Business communication
  * Automated document processing for mid-scale applications
    {% endtab %}

{% tab title="Virtuoso" %}

### **Virtuoso Large**

* **Description:** Our most powerful and versatile general-purpose model, designed to excel at handling complex and varied tasks across domains. With state-of-the-art performance, it offers unparalleled capability for nuanced understanding, contextual adaptability, and high accuracy.
  * **#Parameters:** 72B
  * **Base Model:** Qwen-2.5-72B
  * Open-source and available on Hugging Face under the Apache-2.0 license: [arcee-ai/Virtuoso-Large](https://huggingface.co/arcee-ai/Virtuoso-Large)
* **Top Use Cases:**
  * Advanced content creation, such as technical writing and creative storytelling
  * Data summarization and report generation for cross-functional domains
  * Detailed knowledge synthesis and deep-dive insights from diverse datasets
  * Multilingual support for international operations and communications

### **Virtuoso Medium**

* **Description:** A versatile and powerful model, capable of handling complex and varied tasks with precision and adaptability across multiple domains. Ideal for dynamic use cases requiring significant computational power.
  * **#Parameters:** 32B
  * **Base Model:** Qwen-2.5-32B
  * Open-source and available on Hugging Face under the Apache-2.0 license: [arcee-ai/Virtuoso-Medium-v2](https://huggingface.co/arcee-ai/Virtuoso-Medium-v2)
* **Top Use Cases:**
  * Content generation
  * Knowledge retrieval
  * Advanced language understanding
  * Comprehensive data interpretation

### **Virtuoso Small**

* **Description:** A streamlined version of Virtuoso, maintaining robust capabilities for handling complex tasks across domains while offering enhanced cost-efficiency and quicker response times.
  * **#Parameters:** 14B
  * **Base Model:** Qwen-2.5-14B
  * Open-source and available on Hugging Face under the Apache-2.0 license: [arcee-ai/Virtuoso-Small](https://huggingface.co/arcee-ai/Virtuoso-Small) 
* **Top use cases:** 
  * General-purpose task handling
  * Business communication
  * Automated document processing for mid-scale applications
    {% endtab %}

{% tab title="Maestro" %}

### Maestro

* **Description:** An advanced reasoning model optimized for high-performance enterprise applications. Building on the innovative training techniques first deployed in maestro-7b-preview, Maestro-32B offers significantly enhanced reasoning capabilities at scale, rivaling or surpassing leading models like OpenAI’s O1 and DeepSeek’s R1, but at substantially reduced computational costs.
  * **#Parameters:** 32B
  * **Base Model:** Qwen-2.5-32B
  * Open-source and available on Hugging Face under the Apache-2.0 license: [arcee-ai/Arcee-Maestro-7B-Preview](https://huggingface.co/arcee-ai/Arcee-Maestro-7B-Preview)
  * Hybrid training method:
    1. Warm-up (SFT Phase): Quick supervised fine-tuning phase to prime the model with high-quality reasoning exemplars.
    2. RL Optimization Phase: Utilizes Reinforcement Learning techniques, specifically designed to boost logical coherence, depth of reasoning, and accurate inference by encouraging problem-solving from fundamental principles.
* **Top Use Cases:**
  * Enterprise decision support systems
  * Complex analytical and logical inference tasks
  * Automated research and analysis workflows
  * Generative reasoning for technical and professional contexts
    {% endtab %}

{% tab title="Coder" %}

### Coder Large

* **Description:** A high-performance model tailored for intricate programming tasks, Coder-Large thrives in software development environments. With its focus on efficiency, reliability, and adaptability, it supports developers in crafting, debugging, and refining code for complex systems.
  * **#Parameters:** 32B
  * **Base Model:** Qwen-2.5-32B-Instruct
  * Open-source and available on Hugging Face under the Apache-2.0 license: [arcee-ai/Coder-Large](https://huggingface.co/arcee-ai/Coder-Large)
* **Top use cases:**
  * Writing modular, reusable code across various programming languages
  * Debugging and optimizing performance in large-scale applications
  * Generating efficient algorithms for computationally intensive tasks
  * Supporting DevOps processes, such as script automation and CI/CD pipelines

### Coder Small

* **Description:** A compact, high-performance coding model designed for efficient programming tasks, including generating code, debugging, and optimizing scripts for smaller projects.
  * **#Parameters:** 14B
  * **Base Model:** Qwen-2.5-32B-Instruct
* **Top use cases:**
  * Lightweight development tasks
  * Automated code reviews
  * Generating templates or prototypes quickly, code completion
    {% endtab %}

{% tab title="Caller" %}

### Caller

* **Description:** Engineered for seamless integrations, Caller-Large is a robust model optimized for managing complex tool-based interactions and API function calls. Its strength lies in precise execution, intelligent orchestration, and effective communication between systems, making it indispensable for sophisticated automation pipelines.
  * **#Parameters:** 32B
  * **Base Model:** Qwen-2.5-32B
  * Open-source and available on Hugging Face under the Apache-2.0 license: [arcee-ai/Caller](https://huggingface.co/arcee-ai/Caller)
* **Top use cases:**
  * Managing integrations between CRMs, ERPs, and other enterprise systems
  * Running multi-step workflows with intelligent condition handling
  * Orchestrating external tool interactions like calendar scheduling, email parsing, or data extraction
  * Real-time monitoring and diagnostics in IoT or SaaS environments

{% endtab %}
{% endtabs %}