second_constantine/magistral-small

second_constantine/ magistral-small

434 Downloads Updated 5 months ago

Building upon Mistral Small 3.2 (2506), with added reasoning capabilities, undergoing SFT from Magistral Medium traces and RL on top, it's a small, efficient reasoning model with 24B parameters.(quantized UD-Q5_K_XL)

vision tools thinking 24b

ollama run second_constantine/magistral-small:24b

curl http://localhost:11434/api/chat \
  -d '{
    "model": "second_constantine/magistral-small:24b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='second_constantine/magistral-small:24b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'second_constantine/magistral-small:24b',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code ollama launch claude --model second_constantine/magistral-small:24b

Codex ollama launch codex --model second_constantine/magistral-small:24b

OpenCode ollama launch opencode --model second_constantine/magistral-small:24b

OpenClaw ollama launch openclaw --model second_constantine/magistral-small:24b

Models

View all →

Name

1 model

Size

Context

Input

magistral-small:24b

18GB · 128K context window · Text, Image · 5 months ago

magistral-small:24b

18GB

128K

Text, Image

Readme

Based on the https://huggingface.co/unsloth/Magistral-Small-2509-GGUF

Feature	Value
vision	true (>=0.11.11)
thinking	+/-?
tools	true

Device	Speed, token/s	Context	VRAM, gb	Versions
RTX 3090 24gb	~42	4096	19	UD-Q5_K_XL, 0.12.2
RTX 3090 24gb	~42	15360	21	UD-Q5_K_XL, 0.12.2
M1 Max 32gb	~15	4096	18	UD-Q5_K_XL, 0.12.2
M1 Max 32gb	~15	15360	20	UD-Q5_K_XL, 0.12.2