# Fara 1.5 27B

> Canonical: https://www.overmindlab.ai/models/fara-1-5-27b

27B Qwen 3.5 finetune with a 256K window and tool calling.

| Property | Value |
| --- | --- |
| Model ID | microsoft/Fara1.5-27B |
| Provider | Microsoft |
| Group | Fara |
| Parameters | 27B |
| Tier | Mid |
| Context window | 256K |
| Max training context | 256K |
| Tool calling | Yes |
| Training methods | LoRA |
| Training cost | from $1.22 |
| Serving cost (per 1M output tokens) | from $15.00 |

## Overview

Same architecture as Qwen 3.5 27B: 16 full-attention layers, 256K positions, tool calling. LoRA uses that full window.

Fara 1.5 27B is a Qwen 3.5 27B finetune. Its chat template takes the training path, and a 2-step LoRA completed on the Qwen 3.5 Unsloth image.

## Good for

- Agents that start from the Fara 1.5 checkpoint
- LoRA on a 27B Qwen 3.5 architecture
- Tool-calling rows

Train this model on your agent's data: https://docs.overmindlab.ai/models/training.md