Qwen-Image-Edit-2511
https://huggingface.co/Qwen/Qwen-Image-Edit-2511
import os
import torch
from PIL import Image
from diffusers import QwenImageEditPlusPipeline
pipeline = QwenImageEditPlusPipeline.from_pretrained("Qwen/Qwen-Image-Edit-2511", torch_dtype=torch.bfloat16)
print("pipeline loaded")
pipeline.to('cuda')
pipeline.set_progress_bar_config(disable=None)
image1 = Image.open("input1.png")
image2 = Image.open("input2.png")
prompt = "The magician bear is on the left, the alchemist bear is on the right, facing each other in the central park square."
inputs = {
"image": [image1, image2],
"prompt": prompt,
"generator": torch.manual_seed(0),
"true_cfg_scale": 4.0,
"negative_prompt": " ",
"num_inference_steps": 40,
"guidance_scale": 1.0,
"num_images_per_prompt": 1,
}
with torch.inference_mode():
output = pipeline(**inputs)
output_image = output.images[0]
output_image.save("output_image_edit_2511.png")
print("image saved at", os.path.abspath("output_image_edit_2511.png"))
| Model Name |
Parameter Count |
Disk / Model Size |
Approx. RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
openai/clip-vit-large-patch14 |
~428M |
~1.71 GB |
~2.0 - 2.5 GB |
Zero-shot image classification, text-to-image/image-to-image semantic search, and cross-modal embedding retrieval. |
Not natively supported in Ollama (Vision/Embedding) — Use via transformers / sentence-transformers in Python. |
2. Top Free Open-Weight Models on Hugging Face for Image Tasks
Depending on whether you need Image Generation, Vision-Language Understanding (VLM), Segmentation, or Real-Time Object Detection, here are the absolute best, most liked open models on Hugging Face:
| Model Category |
Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
| Image Generation |
black-forest-labs/FLUX.1-dev |
~12B |
~23.8 GB |
~16 - 24 GB |
State-of-the-art high-fidelity text-to-image synthesis, photorealistic rendering, and typography in images. |
Not supported in Ollama (Use via ComfyUI, DiffusionBee, or diffusers). |
| Fast Image Gen |
black-forest-labs/FLUX.1-schnell |
~12B |
~23.8 GB |
~12 - 16 GB |
Ultra-fast (1–4 steps) real-time text-to-image generation for low-latency pipelines. |
Not supported in Ollama (Use via ComfyUI or diffusers). |
| Classic Image Gen |
stabilityai/stable-diffusion-xl-base-1.0 |
~3.5B |
~6.6 GB |
~8 - 12 GB |
Highly customizable, fine-tuneable text-to-image generation with massive LoRA ecosystem support. |
Not supported in Ollama (Use via Automatic1111 / ComfyUI). |
| Vision Language (VLM) |
Qwen/Qwen2.5-VL-7B-Instruct |
~7B |
~14.5 GB |
~8.0 - 16 GB |
Visual question answering (VQA), reading charts, OCR, document understanding, and analyzing UI screenshots. |
ollama pull qwen2.5-vl |
| Compact VLM |
Qwen/Qwen2.5-VL-3B-Instruct |
~3B |
~6.5 GB |
~4.0 - 6 GB |
Lightweight multimodal chat, image captioning, and structured visual data extraction on consumer hardware. |
ollama pull qwen2.5-vl:3b |
| Promptable Segmentation |
facebook/sam3 (Segment Anything 3) |
~1B |
~4.0 GB |
~6 - 8 GB |
Zero-shot object segmentation and video tracking using text prompts, box prompts, or click points. |
Not supported in Ollama (Use via torch / PyTorch). |
| Real-Time Detection |
roboflow/rf-detr |
~30M - 100M |
~120MB - 400MB |
~0.5 - 2 GB |
End-to-end zero-NMS object detection, instance segmentation, and keypoint tracking for real-time video/CCTV streams. |
Not supported in Ollama (Use via rfdetr Python package). |
| Model Name |
Parameter Count |
Disk / Model Size |
Approx. RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
tencent/HunyuanVideo |
~13B |
~26.0 GB |
~24.0 - 32 GB (FP16) / ~14 - 16 GB (Quantized) |
High-fidelity text-to-video generation with cinematic motion, 720p/1080p resolution, and temporal consistency. |
Not supported in Ollama (Video Diffusion) — Use via Hugging Face diffusers or ComfyUI. |
| Model Category |
Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
| High-Quality Text-to-Video |
tencent/HunyuanVideo |
~13B |
~26.0 GB |
~24.0 - 32 GB |
State-of-the-art text-to-video generation producing cinematic, photo-realistic motion clips. |
Not supported in Ollama (Use via diffusers / ComfyUI) |
| Fast Text-to-Video |
Lightricks/LTX-Video |
~2B |
~4.5 GB |
~8.0 - 12 GB |
Low-latency, memory-efficient real-time text-to-video generation suitable for consumer GPUs. |
Not supported in Ollama (Use via diffusers / ComfyUI) |
| Image & Text-to-Video |
THUDM/CogVideoX-5b |
~5B |
~10.0 GB |
~12.0 - 16 GB |
Highly versatile text-to-video and image-to-video generation for short dynamic animation. |
Not supported in Ollama (Use via diffusers / ComfyUI) |
| Video Understanding & QA |
Qwen/Qwen2.5-VL-7B-Instruct |
~7B |
~14.5 GB |
~8.0 - 16 GB |
Analyzing multi-frame video clips, video chat, OCR in video, dynamic event description, and timestamp detection. |
ollama pull qwen2.5-vl |
| Heavyweight Video LLM |
LMMS-Lab/LLaVA-Video-7B-Qwen2 |
~7B |
~14.0 GB |
~8.0 - 14 GB |
Detailed narrative video comprehension, spatio-temporal reasoning, and long-video analysis. |
ollama pull hf.co/LMMS-Lab/LLaVA-Video-7B-Qwen2-GGUF |
| Video Action Recognition |
MCG-NJU/videomae-base |
~86M |
~340 MB |
~0.5 - 1.0 GB |
Masked autoencoder for efficient video classification, human activity recognition, and action detection. |
Not supported in Ollama (Use via transformers in PyTorch) |
Audio
| Model Name |
Parameter Count |
Disk / Model Size |
Approx. RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
openai/whisper-large-v3 |
~1.55B |
~3.1 GB |
~3.5 - 4.5 GB (FP16) / ~1.5 - 2.0 GB (INT8) |
Multilingual automatic speech recognition (ASR), audio transcription, timestamp generation, and language translation across 100+ languages. |
ollama pull whisper (Available in community Ollama builds & whisper.cpp) |
| Model Category |
Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
| Fast Speech Recognition (ASR) |
openai/whisper-large-v3-turbo |
~809M |
~1.6 GB |
~2.0 - 2.5 GB |
Pruned, 6x faster Whisper variant for real-time transcription and voice-control interfaces. |
ollama pull whisper (or via whisper.cpp / faster-whisper) |
| Ultra-Light Text-to-Speech (TTS) |
hexgrad/Kokoro-82M |
~82M |
~330 MB |
~0.5 - 1.0 GB |
Ultra-fast, lightweight, highly natural text-to-speech generation running efficiently even on low-end CPUs. |
Not supported in Ollama (Use via kokoro Python package or ONNX runtime) |
| Voice Cloning & Expressive TTS |
coqui/XTTS-v2 |
~460M |
~1.8 GB |
~2.0 - 3.0 GB |
Zero-shot voice cloning from a 3-second audio prompt across 17+ languages. |
Not supported in Ollama (Use via TTS library or PyTorch) |
| Full Duplex Audio-to-Audio LLM |
kyutai/moshi |
~7B |
~14.0 GB |
~8.0 - 16 GB |
Real-time bi-directional conversational speech AI supporting simultaneous listening and speaking (low-latency streaming audio). |
Not supported in Ollama (Use via Rust/PyTorch Moshi server) |
| Multimodal Audio LLM |
Qwen/Qwen2-Audio-7B-Instruct |
~7B |
~14.2 GB |
~8.0 - 16 GB |
Direct speech analysis, voice emotion recognition, sound classification, and oral question answering without needing a separate ASR stage. |
ollama pull qwen2-audio |
| Music Generation |
facebook/musicgen-medium |
~1.5B |
~3.2 GB |
~4.0 - 6.0 GB |
Text-conditional music generation (creates instrumental tracks, beats, and background music from text prompts). |
Not supported in Ollama (Use via audiocraft / transformers) |
| Audio & Sound Effects Gen |
facebook/audiogen-medium |
~1.5B |
~3.2 GB |
~4.0 - 6.0 GB |
Text-conditional sound effect synthesis (foley, ambient noise, environmental sounds). |
Not supported in Ollama (Use via audiocraft / transformers) |
| Audio Embedding / Classification |
laion/clap-htsat-fused |
~200M |
~800 MB |
~1.0 - 1.5 GB |
Multimodal zero-shot audio classification, sound search, and audio-text similarity scoring (like CLIP for sound). |
Not supported in Ollama (Use via transformers / msclap) |
Coding
| Model Name |
Parameter Count |
Disk / Model Size |
Approx. RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
Qwen/Qwen2.5-Coder-32B-Instruct |
~32.5B |
~65 GB (FP16) / ~18 - 20 GB (Q4 GGUF) |
~36 - 68 GB (FP16) / ~20 - 24 GB (Q4) |
Full-stack software engineering, complex multi-file code generation, architecture design, bug fixing, and agentic coding matching frontier closed-source models. |
ollama pull qwen2.5-coder:32b |
| Model Category |
Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
| Flagship Code LLM |
Qwen/Qwen2.5-Coder-32B-Instruct |
~32.5B |
~20 GB (Q4) |
~20 - 24 GB (Q4) |
Complex system architecture, multi-file code generation, refactoring, and autonomous agentic coding workflows. |
ollama pull qwen2.5-coder:32b |
| Mid-Sized Workhorse |
Qwen/Qwen2.5-Coder-7B-Instruct |
~7.6B |
~4.7 GB (Q4) |
~5.5 - 8.0 GB |
Outstanding balance of speed and accuracy; handles daily coding, unit testing, and script generation on standard developer laptops. |
ollama pull qwen2.5-coder:7b |
| Ultra-Lightweight / Edge |
Qwen/Qwen2.5-Coder-1.5B-Instruct |
~1.5B |
~1.1 GB (Q4) |
~1.5 - 2.5 GB |
Ultra-fast inline autocompletion, real-time code suggestions, and IDE extensions running on low-spec hardware. |
ollama pull qwen2.5-coder:1.5b |
| MoE High Efficiency |
deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct |
~16B Total (~2.4B Active) |
~9.0 GB (Q4) |
~6.0 - 10.0 GB |
Mixture-of-Experts (MoE) architecture supporting 330+ programming languages with 128k token context window at low inference cost. |
ollama pull deepseek-coder-v2:16b |
| Heavyweight MoE |
deepseek-ai/DeepSeek-Coder-V2-Instruct |
~236B Total (~21B Active) |
~130 GB (Q4) |
~135 - 150 GB |
Frontier-level coding, advanced mathematical reasoning, and repository-scale codebase analysis requiring enterprise server setup. |
ollama pull deepseek-coder-v2:236b |
| Code Completion / FIM |
bigcode/starcoder2-15b |
~15B |
~9.5 GB (Q4) |
~10.0 - 12.0 GB |
Native Fill-in-the-Middle (FIM) code completion, docstring generation, and repository-aware code filling. |
ollama pull starcoder2:15b |
| Lightweight FIM |
bigcode/starcoder2-3b |
~3B |
~2.0 GB (Q4) |
~2.5 - 4.0 GB |
Fast inline Fill-in-the-Middle autocomplete for IDE extensions (VS Code, Neovim) with minimal VRAM footprint. |
ollama pull starcoder2:3b |
| Small Instruction Model |
meta-llama/Llama-3.2-3B-Instruct |
~3.2B |
~2.2 GB (Q4) |
~3.0 - 4.5 GB |
General programming Q&A, bash scripting, regex creation, and lightweight code explanations on edge devices. |
ollama pull llama3.2:3b |
1. General Text Generation & Reasoning (LLMs)
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
Qwen/Qwen2-0.5B |
~490M |
~1.0 GB |
~0.5 - 1.2 GB |
Embedded devices, fast auto-completion, and micro-text generation. |
ollama pull qwen2:0.5b |
meta-llama/Llama-3.2-1B-Instruct |
~1.23B |
~2.5 GB |
~0.8 - 2.5 GB |
Fast, on-device instruction following, local summarization, and mobile chat. |
ollama pull llama3.2:1b |
Qwen/Qwen2.5-1.5B-Instruct |
~1.54B |
~3.1 GB |
~1.0 - 3.5 GB |
Compact local chat, basic structured reasoning, and lightweight workflows. |
ollama pull qwen2.5:1.5b |
meta-llama/Llama-3.2-3B-Instruct |
~3.2B |
~2.2 GB (Q4) |
~3.0 - 4.5 GB |
Mid-sized lightweight general assistant, bash scripting, and document Q&A. |
ollama pull llama3.2:3b |
microsoft/phi-4 |
~14B |
~8.5 GB (Q4) |
~9.0 - 12.0 GB |
High-reasoning small model for math, logic, and multi-turn instruction following. |
ollama pull phi4 |
deepseek-ai/DeepSeek-R1-Distill-Qwen-14B |
~14B |
~9.0 GB (Q4) |
~10.0 - 14.0 GB |
Advanced step-by-step reasoning and mathematical problem-solving. |
ollama pull deepseek-r1:14b |
meta-llama/Llama-3.3-70B-Instruct |
~70B |
~40 GB (Q4) |
~42.0 - 48.0 GB |
High-capability enterprise assistant, complex reasoning, and agentic workflows. |
ollama pull llama3.3:70b |
2. Coding & Software Engineering
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
Qwen/Qwen2.5-Coder-1.5B-Instruct |
~1.5B |
~1.1 GB (Q4) |
~1.5 - 2.5 GB |
Ultra-fast inline IDE autocompletion and low-latency code completion. |
ollama pull qwen2.5-coder:1.5b |
bigcode/starcoder2-3b |
~3B |
~2.0 GB (Q4) |
~2.5 - 4.0 GB |
Fill-in-the-Middle (FIM) code completion and docstring generation. |
ollama pull starcoder2:3b |
Qwen/Qwen2.5-Coder-7B-Instruct |
~7.6B |
~4.7 GB (Q4) |
~5.5 - 8.0 GB |
Daily development, unit testing, bug fixing, and single-file refactoring. |
ollama pull qwen2.5-coder:7b |
deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct |
~16B (~2.4B active) |
~9.0 GB (Q4) |
~6.0 - 10.0 GB |
Mixture-of-Experts (MoE) coding model supporting 330+ languages. |
ollama pull deepseek-coder-v2:16b |
Qwen/Qwen2.5-Coder-32B-Instruct |
~32.5B |
~20 GB (Q4) |
~20.0 - 24.0 GB |
State-of-the-art full-stack software development, architecture, and agentic coding. |
ollama pull qwen2.5-coder:32b |
3. Text Embeddings & Retrieval (RAG / Search)
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
sentence-transformers/all-MiniLM-L6-v2 |
~22.7M |
~90 MB |
~0.25 GB |
Extremely fast dense vector embeddings for low-resource semantic search. |
ollama pull all-minilm |
BAAI/bge-small-en-v1.5 |
~33M |
~130 MB |
~0.3 - 0.5 GB |
Fast, lightweight English embeddings for high-throughput vector databases. |
ollama pull bge-small |
sentence-transformers/all-mpnet-base-v2 |
~109M |
~440 MB |
~0.5 - 1.0 GB |
Standard high-accuracy English sentence embeddings. |
ollama pull hf.co/sentence-transformers/all-mpnet-base-v2 |
nomic-ai/nomic-embed-text-v1.5 |
~137M |
~550 MB |
~0.6 - 1.0 GB |
Long-context (up to 8,192 tokens) text embeddings for document search. |
ollama pull nomic-embed-text |
intfloat/e5-large-v2 |
~335M |
~1.34 GB |
~1.5 - 2.0 GB |
High-precision document retrieval and semantic matching for RAG. |
ollama pull hf.co/intfloat/e5-large-v2 |
BAAI/bge-reranker-v2-m3 |
~568M |
~2.3 GB |
~2.5 - 3.0 GB |
Multilingual cross-encoder candidate re-ranking stage after vector retrieval. |
ollama pull bge-m3 |
4. Vision & Multimodal (Image Tasks)
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
timm/mobilenetv3_small_100.lamb_in1k |
~2.5M |
~10 MB |
< 0.1 GB |
Ultra-light real-time image classification on mobile and edge devices. |
Not supported in Ollama (PyTorch/timm) |
openai/clip-vit-large-patch14 |
~428M |
~1.71 GB |
~2.0 - 2.5 GB |
Zero-shot image classification and cross-modal image-text search. |
Not supported natively in Ollama |
facebook/sam2 |
~1B |
~4.0 GB |
~6.0 - 8.0 GB |
Zero-shot image/video object segmentation and promptable masking. |
Not supported in Ollama (PyTorch) |
Qwen/Qwen2.5-VL-3B-Instruct |
~3B |
~6.5 GB |
~4.0 - 6.0 GB |
Lightweight Vision-Language Model for chart reading, OCR, and image Q&A. |
ollama pull qwen2.5-vl:3b |
stabilityai/stable-diffusion-xl-base-1.0 |
~3.5B |
~6.6 GB |
~8.0 - 12.0 GB |
Open-weight text-to-image synthesis with extensive fine-tuning support. |
Not supported in Ollama (Diffusers) |
Qwen/Qwen2.5-VL-7B-Instruct |
~7B |
~14.5 GB |
~8.0 - 16.0 GB |
Advanced document understanding, screenshot analysis, and visual reasoning. |
ollama pull qwen2.5-vl |
black-forest-labs/FLUX.1-schnell |
~12B |
~23.8 GB |
~12.0 - 16.0 GB |
Ultra-fast (1–4 step) state-of-the-art text-to-image generation. |
Not supported in Ollama (ComfyUI/Diffusers) |
5. Audio, Speech & Sound
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
hexgrad/Kokoro-82M |
~82M |
~330 MB |
~0.5 - 1.0 GB |
Ultra-fast, highly natural text-to-speech (TTS) on low-end CPUs. |
Not supported in Ollama (Python/ONNX) |
laion/clap-htsat-fused |
~200M |
~800 MB |
~1.0 - 1.5 GB |
Zero-shot audio classification, sound search, and audio-text scoring. |
Not supported in Ollama (Transformers) |
coqui/XTTS-v2 |
~460M |
~1.8 GB |
~2.0 - 3.0 GB |
Zero-shot voice cloning from a 3-second sample across 17+ languages. |
Not supported in Ollama (Coqui TTS) |
openai/whisper-large-v3-turbo |
~809M |
~1.6 GB |
~2.0 - 2.5 GB |
Pruned, 6x faster Whisper variant for real-time speech transcription. |
ollama pull whisper |
openai/whisper-large-v3 |
~1.55B |
~3.1 GB |
~3.5 - 4.5 GB |
Multilingual Automatic Speech Recognition (ASR) and translation in 100+ languages. |
ollama pull whisper |
facebook/musicgen-medium |
~1.5B |
~3.2 GB |
~4.0 - 6.0 GB |
Text-conditional music and instrumental track generation. |
Not supported in Ollama (AudioCraft) |
Qwen/Qwen2-Audio-7B-Instruct |
~7B |
~14.2 GB |
~8.0 - 16.0 GB |
Multimodal audio LLM for direct voice reasoning, emotion analysis, and audio Q&A. |
ollama pull qwen2-audio |
6. Video Generation & Understanding
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
MCG-NJU/videomae-base |
~86M |
~340 MB |
~0.5 - 1.0 GB |
Action recognition, video classification, and temporal gesture tracking. |
Not supported in Ollama (PyTorch) |
Lightricks/LTX-Video |
~2B |
~4.5 GB |
~8.0 - 12.0 GB |
Low-latency, memory-efficient real-time text-to-video generation. |
Not supported in Ollama (Diffusers) |
THUDM/CogVideoX-5b |
~5B |
~10.0 GB |
~12.0 - 16.0 GB |
Text-to-video and image-to-video generation for short animations. |
Not supported in Ollama (Diffusers) |
Qwen/Qwen2.5-VL-7B-Instruct |
~7B |
~14.5 GB |
~8.0 - 16.0 GB |
Multi-frame video comprehension, event description, and timestamp detection. |
ollama pull qwen2.5-vl |
tencent/HunyuanVideo |
~13B |
~26.0 GB |
~24.0 - 32.0 GB |
High-fidelity text-to-video generation with cinematic motion and temporal coherence. |
Not supported in Ollama (Diffusers/ComfyUI) |
7. Time-Series Forecasting
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
autogluon/chronos-bolt-small |
~46M |
~180 MB |
~0.3 - 0.5 GB |
Ultra-fast zero-shot time-series forecasting for low-latency production APIs. |
Not supported in Ollama (PyTorch) |
amazon/chronos-2 |
~710M |
~2.8 GB |
~3.0 - 4.0 GB |
Universal probabilistic zero-shot time-series trend forecasting. |
Not supported in Ollama (PyTorch) |
System Monitoring, Log Analysis & Anomaly Detection
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
deepset/bert-base-cased-squad2 |
~110M |
~430 MB |
~0.5 - 1.0 GB |
Extracting key error metrics, IP addresses, and incident parameters from log streams. |
ollama pull hf.co/deepset/bert-base-cased-squad2 |
elastic/distilbert-base-uncased-finetuned-conll03-english |
~66M |
~260 MB |
~0.3 - 0.5 GB |
Named Entity Recognition (NER) for log parsing (parsing user IDs, trace IDs, hostnames). |
ollama pull hf.co/elastic/distilbert-base-uncased-finetuned-conll03-english |
Qwen/Qwen2.5-Coder-1.5B-Instruct |
~1.5B |
~1.1 GB (Q4) |
~1.5 - 2.5 GB |
Fast real-time log parsing, shell script parsing, and lightweight anomaly alerting scripts. |
ollama pull qwen2.5-coder:1.5b |
meta-llama/Llama-3.2-3B-Instruct |
~3.2B |
~2.2 GB (Q4) |
~3.0 - 4.5 GB |
Summarizing system error logs, stack traces, and Kubernetes event messages. |
ollama pull llama3.2:3b |
Qwen/Qwen2.5-7B-Instruct |
~7.6B |
~4.7 GB (Q4) |
~5.5 - 8.0 GB |
Automated root-cause analysis (RCA), SRE incident investigation, and log diagnostic reporting. |
ollama pull qwen2.5:7b |
Finance, Trading & Financial NLP
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
ProsusAI/finbert |
~110M |
~440 MB |
~0.5 - 1.0 GB |
High-throughput financial sentiment analysis on news headlines, SEC filings, and tweets. |
ollama pull hf.co/ProsusAI/finbert |
yiyanghkust/finbert-tone |
~110M |
~440 MB |
~0.5 - 1.0 GB |
Detecting financial tone (positive, negative, neutral) in earnings call transcripts. |
ollama pull hf.co/yiyanghkust/finbert-tone |
FinGPT/fingpt-mt_llama3-8b_lora |
~8B |
~5.0 GB (Q4) |
~6.0 - 8.5 GB |
Financial news analysis, stock sentiment scoring, and market trend summarization. |
ollama pull hf.co/FinGPT/fingpt-mt_llama3-8b_lora |
TheBloke/Fin-Llama2-13B-GGUF |
~13B |
~7.8 GB (Q4) |
~8.5 - 11.0 GB |
Analyzing 10-K/10-Q reports, corporate financial statements, and financial Q&A. |
ollama pull hf.co/TheBloke/Fin-Llama2-13B-GGUF |
Medical, Clinical & Healthcare AI
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
emilyalsentzer/Bio_ClinicalBERT |
~110M |
~440 MB |
~0.5 - 1.0 GB |
Processing clinical notes, discharge summaries, medical entity extraction, and ICD coding. |
ollama pull hf.co/emilyalsentzer/Bio_ClinicalBERT |
openmed-community/granite-4.0-micro-OpenMed |
~3B |
~2.0 GB (Q4) |
~2.5 - 4.0 GB |
Efficient clinical text processing, medical terminology mapping, and lightweight medical Q&A. |
ollama pull hf.co/openmed-community/granite-4.0-micro-OpenMed-GGUF |
BioMistral/BioMistral-7B |
~7B |
~4.4 GB (Q4) |
~5.5 - 8.0 GB |
Biomedical domain question answering, PubMed research paper comprehension, and medical reasoning. |
ollama pull biomistral |
openmed-community/AFM-4.5B-OpenMed-RL-CoT |
~4.5B |
~2.8 GB (Q4) |
~3.5 - 5.0 GB |
Chain-of-Thought medical diagnostic reasoning and clinical trial protocol understanding. |
ollama pull hf.co/openmed-community/AFM-4.5B-OpenMed-RL-CoT-GGUF |
meditron-70b |
~70B |
~40 GB (Q4) |
~42.0 - 48.0 GB |
Enterprise-grade medical literature synthesis, clinical decision support guidance, and medical examination QA. |
ollama pull meditron:70b |
Veda, Upanishads & Sanskrit Literature
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
Dhruvil8/SanskritGPT-Itihasa |
~41.9M |
~170 MB |
~0.2 - 0.4 GB |
Generative experiment modeling classical Sanskrit verse structure, metrical patterns, and epic literature. |
ollama pull hf.co/Dhruvil8/SanskritGPT-Itihasa |
sanganaka/Vedavani-Dataset (ASR Base) |
~39M - 110M |
~200 - 450 MB |
~0.5 - 1.0 GB |
Phonetic speech recognition and chanting alignment for Rig Veda and Atharva Veda verses. |
Not supported in Ollama (PyTorch/Speech) |
Qwen/Qwen2.5-7B-Instruct (Prompt-Guided) |
~7.6B |
~4.7 GB (Q4) |
~5.5 - 8.0 GB |
Translating, explaining, and analyzing Sanskrit philosophical commentaries, Upanishadic dialogues, and Vedic hymns. |
ollama pull qwen2.5:7b |
deepseek-ai/DeepSeek-R1-Distill-Qwen-14B |
~14B |
~9.0 GB (Q4) |
~10.0 - 14.0 GB |
Deep structural analysis of Sanskrit grammar (Panini rules), philosophical comparative studies, and multi-lingual commentary synthesis. |
ollama pull deepseek-r1:14b |
Story Writing for Kids
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
roneneldan/TinyStories-33M |
~33M |
~130 MB |
~0.2 - 0.4 GB |
Generating simple 3- to 4-sentence children's stories using basic vocabulary for early readers. |
ollama pull hf.co/roneneldan/TinyStories-33M |
roneneldan/TinyStories-Instruct-33M |
~33M |
~130 MB |
~0.2 - 0.4 GB |
Instruction-tuned simple story generator for toddlers and pre-k vocabulary learning. |
ollama pull hf.co/roneneldan/TinyStories-Instruct-33M |
meta-llama/Llama-3.2-1B-Instruct |
~1.23B |
~0.8 GB (Q4) |
~1.0 - 1.5 GB |
On-device interactive bedtime story generator with age-appropriate vocabulary controls. |
ollama pull llama3.2:1b |
GhulamMustafa0/children-story-generator |
~7B |
~4.5 GB (Q4) |
~5.5 - 8.0 GB |
Personalized children's story generation based on name, age, mood, and favorite animal. |
ollama pull hf.co/GhulamMustafa0/children-story-generator |
Qwen/Qwen2.5-7B-Instruct |
~7.6B |
~4.7 GB (Q4) |
~5.5 - 8.0 GB |
Creative, multi-chapter children's fiction, moral fables, character-driven adventure stories, and interactive game books. |
ollama pull qwen2.5:7b |
Storybook Generating (Text + Visual / Multimodal Pipeline)
| Model Name |
Parameter Count |
Disk Size |
RAM / VRAM Footprint |
Primary Use Case / Application |
Ollama Pull Command / Availability |
Pipeline Step 1 (Story & Scene Descriptions): meta-llama/Llama-3.2-3B-Instruct |
~3.2B |
~2.2 GB (Q4) |
~3.0 - 4.5 GB |
Writes the story text and crafts consistent image prompts page-by-page. |
ollama pull llama3.2:3b |
Pipeline Step 2 (Illustration - Lightweight): stabilityai/stable-diffusion-2-1-base |
~1.0B |
~3.4 GB |
~4.0 - 6.0 GB |
Generates colorful, stylized illustrations for each storybook page. |
Not supported in Ollama (Diffusers) |
Pipeline Step 2 (Illustration - High Quality): black-forest-labs/FLUX.1-schnell |
~12B |
~23.8 GB |
~12.0 - 16.0 GB |
Ultra-fast high-fidelity digital art and character-consistent picture book illustrations. |
Not supported in Ollama (ComfyUI/Diffusers) |
Pipeline Step 3 (Layout / OCR Verification): Qwen/Qwen2.5-VL-3B-Instruct |
~3B |
~2.0 GB (Q4) |
~4.0 - 6.0 GB |
Inspects generated storybook pages to verify visual layout, text alignment, and aesthetic composition. |
ollama pull qwen2.5-vl:3b |