Skip to content
Computer vision

Krypton Fly 3.0

A multimodal model focused on deep reasoning

2 min read

Krypton Fly 3.0 is a standalone universal multimodal AI model by MyJellyLab with 9.4 billion parameters. The model was trained using QwQ-32B reasoning distillation and optimization methods such as NEFTune and Constitutional DPO. It delivers analytical rigor, high-quality performance with the Linux terminal and programming, and fluent English and Hindi proficiency alongside native image and video understanding. This new release is well-suited for developers, researchers, and hobbyists looking to run local experiments and tackle complex tasks.

Key features

  • Multimodality: Support for text, images, and video.
  • Reasoning distillation: Utilization of OpenThoughts3-1.2M approaches for deep analysis.
  • Engineering tasks: Reliable performance with Linux, the terminal, and verified code.
  • Bilingual capabilities: High response accuracy in English and Hindi.

What you can do

  • Development and debugging: Use the model for writing and reviewing code, as well as terminal workflows.
  • Media analysis: Parse images and video materials within local projects.
  • Research: Conduct deep information synthesis and solve tasks that require step-by-step reasoning.
  • Bot integration: Deploy the model as a smart assistant for English- and Hindi-speaking audiences.

Example

import torch
from transformers import AutoProcessor, Qwen3_5ForConditionalGeneration

model_id = "myjellylabs/Krypton-Fly-3.0"
processor = AutoProcessor.from_pretrained(model_id)
model = Qwen3_5ForConditionalGeneration.from_pretrained(
    model_id,
    dtype=torch.float16,
    device_map="auto",
)

FRONTIER_SYSTEM_PROMPT = (
    "You are Krypton Fly 3.0, developed by MyJellyLab. Reason from first principles, "
    "verify edge cases and trade-offs, be candid about uncertainty without fabricating facts, "
    "and deliver structured, high-density answers."
)
messages = [
    {"role": "system", "content": FRONTIER_SYSTEM_PROMPT},
    {"role": "user", "content": "Who are you?"},
]
prompt = processor.tokenizer.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True, enable_thinking=True
)
inputs = processor.tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=1024)
print(processor.tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=False))
Read moreDetails
• Architecture: Multimodal 9.4B parameter model
• Requirements: transformers library >= 5.6.0, dtype=torch.float16 support
• License: Apache-2.0
• Limitations: Possible hallucinations on rare facts, weak baseline content filtering
Discuss Telegram 1

Will it run?

Runs in F16

Fits without quantisation.

  • F16weights 18 GB · ready files20 GB
  • FP8 / INT8recommendedweights 8.8 GB10 GB
  • 4-bitweights 4.9 GB5.9 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2080 Ti · F16 — from RTX 4090

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # myjellylabs/Krypton-Fly-3.0
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download myjellylabs/Krypton-Fly-3.0 --local-dir ~/models/Krypton-Fly-3.0
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Krypton-Fly-3.0
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of Krypton-Fly-3.0 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 2All on Hugging Face

About the model

myjellylabs/Krypton-Fly-3.0
Author
myjellylabs
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Computer vision

Qwen3.81-27B

Qwen3.81-27B