# iFreeLume AI & Privacy Developer Tools (tool.ifreelume.com)

> Free, 100% browser-local AI engineering calculators (LLM GPU VRAM calculator, LLM Token & API pricing comparison, AI Markdown to Word cleaner), privacy media converters (WebP/AVIF/PNG/JPG/SVG/EXIF stripper), and developer utilities.

## Core Interactive Tools
- [LLM VRAM & GPU Calculator](https://tool.ifreelume.com/en/tools/llm-vram-calculator): Calculate exact GPU VRAM, RAM & bandwidth needed to run DeepSeek-R1, Llama 3.3, Qwen 2.5 locally (FP16 / GGUF Q4_K_M / Q8 / vLLM).
- [LLM Token & API Pricing Calculator](https://tool.ifreelume.com/en/tools/llm-api-pricing-calculator): Count tokens in real time and compare API costs across DeepSeek-R1/V3, GPT-4o, OpenAI o3-mini, Claude 3.7 Sonnet & Gemini 2.0.
- [AI Markdown to Word / Rich Text Cleaner](https://tool.ifreelume.com/en/tools/ai-markdown-cleaner): Strip Markdown symbols (**bold**, ### headers), fix LaTeX math, and copy clean formatted rich text directly into Word, Notion, or WeChat.
- [Local Privacy Image Converter & Compressor](https://tool.ifreelume.com/en/tools/image-converter): Batch convert WebP, AVIF, PNG, JPG, SVG, BMP & compress images 100% locally in your browser. Zero server uploads.
- [Photo EXIF & GPS Metadata Remover](https://tool.ifreelume.com/en/tools/exif-remover): Inspect hidden camera/GPS metadata in photos and strip all EXIF, IPTC, XMP tags locally before sharing online.
- [JSON Formatter, Validator & Struct Generator](https://tool.ifreelume.com/en/tools/json-formatter): Format, minify, validate, sort keys, and convert JSON payloads into Go Structs, TypeScript Interfaces, or Python Pydantic models.
- [cURL to Python, Go, Node.js & PHP Converter](https://tool.ifreelume.com/en/tools/curl-to-code): Convert browser DevTools 'Copy as cURL' commands into clean Python requests, Go net/http, Node.js fetch, or PHP code.
- [Client-Side JWT Decoder & Debugger](https://tool.ifreelume.com/en/tools/jwt-decoder): Decode JWT Header and Payload locally, check token expiration (exp/iat/nbf) countdown, and verify HMAC-SHA256 signatures.
- [Cron Expression Generator & Schedule Predictor](https://tool.ifreelume.com/en/tools/cron-generator): Parse 5-field Linux Crontab expressions into plain English/Chinese and preview the exact next 10 execution timestamps.
- [Base64, SHA-256 Hash, Timestamp, UUID & UTM Builder](https://tool.ifreelume.com/en/tools/encode-hash-uuid): All-in-one developer swiss-army knife: UTF-8 Base64, URL encode, SHA-256/512, Unix timestamp converter, bulk UUID v4 & UTM link builder.

## Open-Weight LLM GPU VRAM Calculators
- [DeepSeek-R1 671B (MoE) (671B) VRAM Calculator](https://tool.ifreelume.com/en/models/deepseek-r1-671b-vram-calculator): Flagship reasoning MoE model (37B active parameters). Requires multi-GPU server or high-memory Mac/CPU+GPU hybrid setup.
- [DeepSeek-R1-Distill-Llama-70B (70.6B) VRAM Calculator](https://tool.ifreelume.com/en/models/deepseek-r1-distill-llama-70b-vram-calculator): Top-tier distilled reasoning model rivaling o1-mini. Runs great in Q4_K_M on Dual RTX 3090/4090 (48GB) or Mac 64GB.
- [DeepSeek-R1-Distill-Qwen-32B (32.8B) VRAM Calculator](https://tool.ifreelume.com/en/models/deepseek-r1-distill-qwen-32b-vram-calculator): Sweet-spot local reasoning model! Fits inside a single 24GB GPU (RTX 3090 / RTX 4090) using Q4_K_M quantization.
- [DeepSeek-R1-Distill-Qwen-14B (14.8B) VRAM Calculator](https://tool.ifreelume.com/en/models/deepseek-r1-distill-qwen-14b-vram-calculator): Ideal mid-size reasoning model for 12GB–16GB GPUs (RTX 4070 Ti, RTX 4080, RTX 4060 Ti 16GB, Mac 16GB).
- [DeepSeek-R1-Distill-Llama-8B (8.03B) VRAM Calculator](https://tool.ifreelume.com/en/models/deepseek-r1-distill-llama-8b-vram-calculator): Fast, lightweight reasoning model that fits comfortably on 8GB consumer GPUs (RTX 3060, RTX 4060) in Q4_K_M.
- [DeepSeek-R1-Distill-Qwen-7B (7.62B) VRAM Calculator](https://tool.ifreelume.com/en/models/deepseek-r1-distill-qwen-7b-vram-calculator): Strong math and Chinese/English reasoning for 6GB–8GB GPUs and 8GB/16GB Apple Silicon laptops.
- [DeepSeek-R1-Distill-Qwen-1.5B (1.54B) VRAM Calculator](https://tool.ifreelume.com/en/models/deepseek-r1-distill-qwen-1-5b-vram-calculator): Ultra-lightweight edge reasoning model (<1.5GB VRAM in Q4_K_M). Runs on any laptop, CPU, or mobile device.
- [DeepSeek-V3 671B (MoE) (671B) VRAM Calculator](https://tool.ifreelume.com/en/models/deepseek-v3-671b-vram-calculator): 671B MoE general-purpose model with 37B active parameters per token.
- [Qwen-2.5-72B-Instruct (72.7B) VRAM Calculator](https://tool.ifreelume.com/en/models/qwen-2-5-72b-vram-calculator): Alibaba's flagship 72B multilingual & coding powerhouse. Requires ~45GB VRAM in Q4_K_M.
- [Qwen-2.5-32B-Instruct (32.5B) VRAM Calculator](https://tool.ifreelume.com/en/models/qwen-2-5-32b-vram-calculator): Fits in a single 24GB GPU (RTX 3090/4090) at Q4_K_M with near-72B performance.
- [Qwen-2.5-Coder-32B-Instruct (32.5B) VRAM Calculator](https://tool.ifreelume.com/en/models/qwen-2-5-coder-32b-vram-calculator): State-of-the-art open-source coding LLM for local Copilot/Cursor integration on 24GB GPUs.
- [QwQ-32B (Reasoning) (32.5B) VRAM Calculator](https://tool.ifreelume.com/en/models/qwq-32b-vram-calculator): Qwen's dedicated 32B reasoning and reinforcement-learning model designed for single 24GB GPUs.
- [Qwen-2.5-14B-Instruct (14.7B) VRAM Calculator](https://tool.ifreelume.com/en/models/qwen-2-5-14b-vram-calculator): Runs smoothly on 12GB–16GB GPUs in Q4_K_M/Q5_K_M with 128K context support.
- [Qwen-2.5-7B-Instruct (7.61B) VRAM Calculator](https://tool.ifreelume.com/en/models/qwen-2-5-7b-vram-calculator): Fast 7B model requiring only ~5.2GB VRAM in Q4_K_M, great for 8GB RTX GPUs.
- [Llama-3.3-70B-Instruct (70.6B) VRAM Calculator](https://tool.ifreelume.com/en/models/llama-3-3-70b-vram-calculator): Meta's latest 70B model matching Llama-3.1-405B quality at a fraction of the VRAM cost (~43GB in Q4_K_M).
- [Llama-3.1-405B-Instruct (405B) VRAM Calculator](https://tool.ifreelume.com/en/models/llama-3-1-405b-vram-calculator): Massive 405B dense frontier model. Needs ~235GB VRAM in Q4_K_M (e.g., 4x A100 80GB or 8x RTX 4090).
- [Llama-3.1-8B-Instruct (8.03B) VRAM Calculator](https://tool.ifreelume.com/en/models/llama-3-1-8b-vram-calculator): The industry standard 8B model. Requires ~5.6GB VRAM in Q4_K_M, ideal for any 8GB GPU.
- [Llama-3.2-11B-Vision-Instruct (10.7B) VRAM Calculator](https://tool.ifreelume.com/en/models/llama-3-2-11b-vision-vram-calculator): Multimodal vision-language model for local OCR, chart reading, and image understanding (~7.5GB in Q4_K_M).
- [Llama-3.2-3B-Instruct (3.21B) VRAM Calculator](https://tool.ifreelume.com/en/models/llama-3-2-3b-vram-calculator): Compact 3B model needing only ~2.4GB VRAM in Q4_K_M. Perfect for laptops and edge devices.
- [Gemma-2-27B-IT (27.2B) VRAM Calculator](https://tool.ifreelume.com/en/models/gemma-2-27b-vram-calculator): Google's 27B model engineered to fit inside a single 24GB GPU (RTX 3090/4090) in Q4_K_M/Q5_K_M.
- [Gemma-2-9B-IT (9.24B) VRAM Calculator](https://tool.ifreelume.com/en/models/gemma-2-9b-vram-calculator): Outperforms most 8B-10B models in creative writing and instruction following (~6.4GB VRAM in Q4_K_M).
- [Mistral-Large-2-123B (123B) VRAM Calculator](https://tool.ifreelume.com/en/models/mistral-large-2-123b-vram-calculator): 123B multilingual and coding giant. Needs ~74GB VRAM in Q4_K_M (fits on 96GB/128GB Mac Studio or 4x24GB GPUs).
- [Mixtral-8x22B-Instruct (MoE) (141B) VRAM Calculator](https://tool.ifreelume.com/en/models/mixtral-8x22b-vram-calculator): 141B sparse MoE model (39B active parameters) offering fast inference when loaded across multi-GPU setups.
- [Mixtral-8x7B-Instruct (MoE) (46.7B) VRAM Calculator](https://tool.ifreelume.com/en/models/mixtral-8x7b-vram-calculator): Classic 46.7B MoE model (12.9B active). Requires ~28GB VRAM in Q4_K_M.
- [Mistral-Nemo-12B-Instruct (12.2B) VRAM Calculator](https://tool.ifreelume.com/en/models/mistral-nemo-12b-vram-calculator): Co-developed by Mistral & NVIDIA for 12GB GPUs (RTX 3060 12GB / RTX 4070). Takes ~8.1GB in Q4_K_M.
- [Microsoft Phi-4 14B (14.7B) VRAM Calculator](https://tool.ifreelume.com/en/models/phi-4-14b-vram-calculator): Microsoft's synthetic-data trained 14B powerhouse excelling at STEM, math, and logic on 12GB–16GB GPUs.
- [Cohere Command R+ 104B (104B) VRAM Calculator](https://tool.ifreelume.com/en/models/command-r-plus-104b-vram-calculator): Enterprise RAG & tool-use specialist (104B parameters, ~62GB VRAM in Q4_K_M).
- [Yi-1.5-34B-Chat (34.4B) VRAM Calculator](https://tool.ifreelume.com/en/models/yi-1-5-34b-vram-calculator): Strong bilingual 34B model that fits inside 24GB VRAM using Q4_K_M quantization.
- [GLM-4-9B-Chat (9.4B) VRAM Calculator](https://tool.ifreelume.com/en/models/glm-4-9b-vram-calculator): Zhipu AI's 9B model with ultra-efficient MQA KV cache, enabling huge context windows on 8GB–12GB GPUs.
- [InternLM2.5-20B-Chat (19.9B) VRAM Calculator](https://tool.ifreelume.com/en/models/internlm2-5-20b-vram-calculator): 20B reasoning and tool-calling model fitting comfortably on 16GB–24GB GPUs (~12.8GB in Q4_K_M).
- [StarCoder2-15B (15.7B) VRAM Calculator](https://tool.ifreelume.com/en/models/starcoder2-15b-vram-calculator): Trained on 600+ programming languages, ideal for local code completion on 12GB GPUs.
- [Mistral Codestral-22B (22.2B) VRAM Calculator](https://tool.ifreelume.com/en/models/codestral-22b-vram-calculator): 22B coding specialist supporting 80+ languages (~14GB VRAM in Q4_K_M, runs on 16GB/24GB GPUs).
- [FLUX.1-dev 12B (Image Diffusion) (12B) VRAM Calculator](https://tool.ifreelume.com/en/models/flux-1-dev-12b-vram-calculator): 12B rectified flow image generator. Needs ~24GB VRAM in FP16/BF16 or ~10GB–12GB in GGUF Q4/FP8.
- [Stable Diffusion 3.5 Large (8B) (8.1B) VRAM Calculator](https://tool.ifreelume.com/en/models/sd3-5-large-8b-vram-calculator): 8.1B MMDiT image synthesis model. Fits on 16GB GPUs in FP16 or 8GB–12GB GPUs in FP8/Q4.
- [Wan2.1-T2V-14B (Video Generation) (14.2B) VRAM Calculator](https://tool.ifreelume.com/en/models/wan2-1-14b-vram-calculator): Open-source 14B text/image-to-video model. Runs on 16GB–24GB GPUs using FP8/GGUF quantization + offloading.
- [Gemma-3-27B-IT (Multimodal) (27.4B) VRAM Calculator](https://tool.ifreelume.com/en/models/gemma-3-27b-vram-calculator): Google's 128K-context multimodal 27B model engineered to run on a single 24GB GPU in Q4_K_M.