Kex 1.5 is Kexity AI’s second generation of flagship TLMs for efficient local inference. Kex 1.5 supports tool calling, thinking, and features token-efficient thinking and reasoning for compute-constrained environments.

Use Case

This model is for customers with extremely constrained compute or low-latency applications. Kex 1.5 punches above it’s weight in agentic use-cases, and is useful for tasks such as the following:

• Agents running on edge/IoT devices with less than 1 GB of RAM
• Low-latency chatbots and agents for environments where speed matters

Speed Comparison

Examples

1. Python

Quick example of a Kex 1.5 tool calling setup

# example usage:
# Prompt: What is the weather today?
# Calling: get_weather({'city': 'paris'})
# Response: It's a sunny 22°C in Paris. Would you like advice on how to protect yourself outdoors?

import json
from rich import print
from ollama import chat
model = 'KexityAI/kex1.5'

def get_weather(city: str) -> str:
  return json.dumps({'city': city, 'temperature': 22, 'unit': 'celsius', 'condition': 'sunny'})

messages = [{"role": "system", "content": "You are a helpful general purpose assistant. The user is located in Paris."}]

while True:
  prompt = input("Prompt: ")

  if prompt == "":
    break

  messages.append({"role": "user", "content": prompt})

  response = chat(model, messages=messages, tools=[get_weather])

  if response.message.tool_calls:
    tool = response.message.tool_calls[0]
    print(f'Calling: {tool.function.name}({tool.function.arguments})')

    result = get_weather(**tool.function.arguments)

    messages.append(response.message)
    messages.append({'role': 'tool', 'content': result})

    final = chat(model, messages=messages)
    print('Response:', final.message.content)
  else:
    print('Response:', response.message.content)

2. Quick on-device AI on free CPU Colab instances

Kex 1.5 is light enough to run on a CPU-only Colab instance!

Cell 1:

try:
  import ollama
except:
  import os, time, threading
  !apt-get install zstd
  !curl -fsSL https://ollama.com/install.sh | sh
  !pip install ollama
  def worker():
    os.system("ollama serve")
  threading.Thread(target=worker).start()
  time.sleep(1)
  !ollama pull KexityAI/kex1.5

Cell 2:

import os, threading, time

def worker():
  os.system("ollama serve")

threading.Thread(target=worker).start()

time.sleep(1)