Fine-tuned 3B coding specialist optimized for Claude Code CLI and agentic tool use. Runs 100% locally in NVIDIA GPU VRAM (GTX 1650 4GB compatible).

tools

ollama run hisparshmishra1/claude-code-3b

curl http://localhost:11434/api/chat \
  -d '{
    "model": "hisparshmishra1/claude-code-3b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='hisparshmishra1/claude-code-3b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'hisparshmishra1/claude-code-3b',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 1 week ago

1 week ago

cfa490daf50c · 1.9GB

model

archqwen2

parameters3.09B

quantizationQ4_K_M

1.9GB

template

{{- if .Suffix }}<|fim_prefix|>{{ .Prompt }}<|fim_suffix|>{{ .Suffix }}<|fim_middle|> {{- else if .M

1.6kB

license

Qwen RESEARCH LICENSE AGREEMENT Qwen RESEARCH LICENSE AGREEMENT Release Date: September 19, 2024 By

7.4kB

system

You are Claude Code, Anthropic's official CLI for software development. Available tools: Bash, Read,

1.1kB

params

{ "num_batch": 512, "num_ctx": 32768, "num_gpu": 99, "stop": [ "<|im_start|>

160B

Readme

Claude Code 3B Specialist (`hisparshmishra1/claude-code-3b`)

A specialized, ultra-efficient 3B parameter coding model based on Qwen 2.5 Coder 3B Instruct, fine-tuned and configured specifically for Claude Code CLI, autonomous coding agents, and agentic tool execution.

Designed to run 100% resident in GPU VRAM on entry-level hardware like the NVIDIA GeForce GTX 1650 (4 GB VRAM) with zero CPU spillover.

Key Highlights

Tailored for Claude Code: System prompts and stop sequences engineered specifically for Claude Code agentic tools (Bash, Edit, Read, Write, Glob, Grep).
100% GPU VRAM Resident: Uses ~1.9 GB for model weights and fits comfortably in 4 GB VRAM GPUs with room for KV cache.
Fast Local Inference: Generates surgical code diffs and responses in seconds.
Quantization: High-quality Q4_K_M GGUF quantization.

Quickstart

1. Run in Terminal (Direct Chat)

”`bash ollama run hisparshmishra1/claude-code-3b