LLM Tools

Tools for working with large language models: token counters, cost calculators, prompt analyzers, and model comparison utilities. Every tool runs in your browser; pasted text never leaves your device.

All LLM Tools

• OpenAI Token Counter Count tokens for GPT-4o, GPT-4, GPT-3.5, o1, o3 with cost estimates.
• Claude Token Counter Count tokens for Claude 3.5 Sonnet, Opus, Haiku, and Claude 4.x.
• Gemini Token Counter Count tokens for Gemini 1.5 Pro, Flash, and Gemini 2.x.
• LLM API Cost Calculator Estimate per-call and monthly costs across LLM providers.
• Multi-Model Token Comparison Compare token counts and costs across GPT, Claude, Gemini, Llama side by side.
• Embedding Cost Calculator Estimate embedding API costs across providers and dimensions.
• System Prompt Analyzer Analyze prompt length, structure, and estimated cost per call.
• Prompt Token Cost Calculator Compare what one prompt costs across every major model. Cached vs uncached pricing.
• ChatGPT / Claude Chat Exporter Format a ChatGPT or Claude conversation export as Markdown, JSON, or HTML.
• OpenAI Pricing Calculator Estimate OpenAI API cost for any GPT model. Per-call + monthly projection.
• Claude Pricing Calculator Estimate Anthropic Claude API cost. Opus / Sonnet / Haiku with prompt caching.
• DeepSeek Token Counter Count tokens for DeepSeek-V3 and DeepSeek-R1 with cost estimates.
• OpenAI Function / Tool Schema Builder Build OpenAI function-calling JSON schemas visually. Parameter types, descriptions, required.
• Anthropic Tool Definition Builder Build Anthropic Claude tool definitions visually. JSON schema + description.
• Recursive Text Chunker (RAG) Split long text into overlapping chunks for embeddings. Configurable size and overlap.
• Embedding Model Comparison Table Compare every major embedding model: dimensions, max input, price, MTEB score.
• Cosine Similarity Calculator Compute cosine similarity between two vectors. Paste embedding vectors, get the score.
• RAG Cost Estimator Estimate end-to-end RAG cost: embedding + storage + retrieval + LLM completion.
• Vector DB Pricing Calculator Compare Pinecone / Qdrant / Weaviate / pgvector for your vector count and queries.
• JSON Extractor from LLM Output Pull JSON objects out of mixed-content LLM responses (with code fences, prose, retries).
• Code Block Extractor Extract every fenced code block from Markdown / LLM output. Group by language.
• Prompt PII Redactor Redact emails, phone numbers, names, addresses from text before sending to an LLM.
• Prompt Injection Scanner Scan user input for common prompt-injection patterns: instruction overrides, jailbreaks.
• GPT-5 Token Counter Token counter dedicated to GPT-5 / GPT-5 Mini / Nano. Uses o200k_base tokenizer.
• GPT-4o Token Counter Token counter dedicated to GPT-4o. Uses o200k_base tokenizer.
• GPT-3.5 Token Counter Token counter dedicated to GPT-3.5 Turbo. Uses cl100k_base tokenizer.
• Llama Token Counter Estimate tokens & cost for Llama 3.1 / 3.3 (8B, 70B, 405B). 128K tiktoken vocab.
• Mistral Token Counter Estimate tokens & cost for Mistral Large, Small, Nemo & Codestral. Tekken tokenizer.
• Qwen Token Counter Estimate tokens & cost for Qwen2.5 (7B–72B & Coder). ~152K BPE vocab.
• Gemma Token Counter Estimate tokens & cost for Gemma 2 (27B, 9B, 2B). 256K vocab, 8K context.
• Grok Token Counter Estimate tokens & cost for xAI Grok 2, Grok 2 mini & Grok Beta. 131K window.
• LLM Context Window Comparison Compare context windows of GPT, Claude, Gemini, Llama & more. See how much text fits.
• AI Image Generation Cost Calculator Compare cost of DALL·E, GPT-image, Flux, Imagen & Stable Diffusion per image.
• Self-Hosting vs API Cost Calculator Compare self-hosting an open LLM on GPUs against per-token API pricing. Find break-even.
• Transformer Parameter Calculator Count LLM parameters from architecture: vocab, hidden size, layers, heads, GQA, FFN.
• LLM Training FLOPs Calculator Estimate training/inference FLOPs (6ND), GPU-hours, wall-clock time & cost.
• GPU Cloud Price Comparison Compare H100, A100, 4090 rental prices across AWS, Lambda, RunPod, Vast & more.
• Batch API & Caching Savings Calculator Calculate savings from the Batch API (50% off) and prompt caching for OpenAI, Claude, Gemini.
• Prompt Anonymizer — Replace Names & Entities Replace person names, company names, product codes with anonymous placeholders.
• Batch Prompt Formatter Apply one prompt template across CSV rows. Generates the batch as JSON for API calls.
• Prompt to Python Script Wrap a prompt into a runnable Python script (OpenAI / Anthropic SDK).
• Prompt to JavaScript / Node Script Wrap a prompt into a runnable Node script (OpenAI / Anthropic SDK).
• Prompt to cURL Wrap a prompt into a working cURL command for any major LLM API.
• Conversation Log Parser Parse any chat transcript into structured turns. Works with copy-pasted UIs.
• Multi-Shot to Single-Shot Converter Collapse a few-shot prompt with examples into one big single-shot prompt.
• Persona / Role Prompt Generator Build system prompts with expertise, tone, format, refusal patterns from a template.
• Negative Prompt Builder (Image Gen) Build negative prompts for Stable Diffusion / Midjourney / Flux from category tags.
• Output Schema → Prompt Instruction Convert a JSON Schema into the prompt text that asks an LLM to produce matching JSON.
• Prompt Diff — For Fine-Tuning Datasets Diff two prompts side-by-side. Highlight added/removed instructions, examples, tokens.
• Gemini Function Declaration Builder Build Gemini FunctionDeclaration JSON schemas. Compatible with Google AI Studio + Vertex.
• Embedding Dimension Calculator (Storage + Memory) Calculate storage and RAM for an embedding index by dimension and document count. Compare float32 vs int8 quantization.
• LLM Cost Comparator (Cross-Provider) Compare API cost across providers for the same input/output token count. OpenAI, Anthropic, Google, DeepSeek, Mistral.
• Tokens-per-Word Chart by Model Visualise how each tokenizer (GPT-4o, Claude, Gemini, Llama) splits English / multilingual text. Compare token efficiency.
• LLM Structured Output Validator (JSON Schema) Validate LLM responses against a JSON Schema. Catch missing fields, wrong types, format errors before sending to your application.
• Prompt Injection Tester (LLM Vulnerability Probe) Test an LLM prompt against 30+ known injection patterns. Helps harden system prompts before production.
• LLM VRAM Calculator Calculate the GPU VRAM needed to run any LLM at a given quantization and context length, and see which GPUs fit.
• GGUF VRAM & Size Calculator Estimate the file size and VRAM for any GGUF quant (Q2_K to Q8_0, F16) of a model, including KV cache.
• Chat Template Formatter Wrap a conversation into a model's exact prompt format — ChatML, Llama 3, Mistral, Gemma, Phi-3, Alpaca, Vicuna — with every special token shown.
• LLM Download Size Calculator Estimate an LLM download size from parameters and GGUF quant (Q4_K_M, Q8_0, F16) using real bits-per-weight.
• MergeKit Config Generator Generate a valid mergekit YAML for SLERP, TIES, DARE, linear, task-arithmetic or passthrough model merges.
• GGUF Metadata Reader Drop a .gguf file and read its header — architecture, context length, layers, quantization, tokenizer. Parsed in your browser, nothing uploaded.
• RoPE / YaRN Context Extension Calculator Work out the RoPE scaling factor to extend a model context (linear, dynamic-NTK, YaRN) and get the rope_scaling config for transformers.
• Tokens per Second Calculator Estimate LLM inference speed in tokens per second from your GPU memory bandwidth and the model size — the memory-bound limit.
• Stable Diffusion VRAM Calculator Estimate VRAM for Stable Diffusion, SDXL, and Flux image generation by model, precision, resolution, and batch size.
• LLM GPU Requirements Checker Check whether a specific LLM fits on your GPU at a chosen quantization and context, with offload and multi-GPU advice.
• KV Cache Size Calculator Calculate transformer KV cache memory from layers, heads, context length, batch size, and precision — the hidden VRAM cost of long context.
• LLM Fine-Tuning Cost Calculator Estimate the GPU-hours and dollar cost to fine-tune an LLM with LoRA, QLoRA, or full fine-tuning across cloud GPUs.
• Prompt Caching Savings Calculator Calculate how much prompt caching saves on Anthropic and OpenAI API bills given your cached tokens, hit rate, and request volume.
• Vision Token Calculator Calculate how many tokens an image costs on GPT-4o, Claude, and Gemini from its dimensions, plus the dollar cost.
• Ollama Modelfile Generator Build a valid Ollama Modelfile from a form — base model, system prompt, parameters, template, and stop tokens.
• ChatML Formatter Convert a conversation into ChatML with im_start and im_end tokens, the chat format used by many open models.
• Safetensors Metadata Viewer Inspect a .safetensors file header locally — tensor names, shapes, dtypes, and total parameter count — without uploading.
• LoRA Memory Calculator Estimate LoRA fine-tuning trainable parameters and adapter, gradient and optimizer memory from model dimensions, rank and target modules. GQA-aware, with presets.
• BLEU and ROUGE Calculator Compute BLEU-1 to BLEU-4 with brevity penalty and ROUGE-1/2/L precision, recall and F1 between a candidate and reference text. Runs in your browser.
• Confusion Matrix Calculator Build a confusion matrix from actual,predicted pairs with accuracy, per-class precision, recall and F1, plus macro, micro and weighted averages.
• Perplexity Calculator Calculate language-model perplexity from per-token log-probabilities, with cross-entropy in nats and bits. Supports natural-log, log2 and log10 inputs.
• MoE Active Parameters Calculator Work out total versus active per-token parameters of a Mixture-of-Experts model from expert count, top-k, params per expert and shared layers. Presets included.
• Transformer FLOPs Calculator Estimate transformer forward and training FLOPs from sequence length, dimension, layers, FFN size and vocab, with an attention versus MLP breakdown.
• LLM Sampling Visualizer See how temperature, top-k and top-p reshape a model next-token distribution. Enter logits and watch the probabilities sharpen, flatten and truncate live.
• Logprobs Visualizer Paste OpenAI-style logprobs JSON for a colour-coded confidence heatmap, per-token probabilities, perplexity and the lowest-confidence tokens.
• Fine-tuning JSONL Validator Validate a chat or prompt/completion JSONL fine-tuning dataset — JSON, roles, structure, duplicates, missing assistant turns — and estimate token count.
• OpenAI API Error Code Explainer Look up any OpenAI API error — 400, 401, 403, 429, 500, 503 and insufficient_quota — with the cause in plain English and the exact fix. Covers rate limits, context-length and auth errors.
• Anthropic / Claude API Error Explainer Decode Anthropic and Claude API errors — the difference between 429 rate_limit and 529 overloaded, plus 400, 401, 403, 413 and 500 — with the cause and the right retry or backoff.
• LLM SSE Stream Decoder Paste a raw Server-Sent Events stream from an LLM API and see each data: chunk parsed, deltas concatenated and the final assembled message. Handles OpenAI and Anthropic event formats.
• OpenAI Batch API JSONL Builder Build and validate a JSONL file for the OpenAI Batch API — one request per line with custom_id, method, url and body — and catch malformed lines before you upload.
• TPM / RPM Rate Limit Calculator Work out whether your workload fits an OpenAI or Anthropic tier — convert requests and tokens per minute into the TPM/RPM headroom you need, and find the concurrency that stays under the cap.
• ShareGPT ↔ OpenAI Dataset Converter Convert fine-tuning datasets between ShareGPT (conversations/from/value) and OpenAI (messages/role/content) formats, both directions, with role mapping and validation.
• Alpaca ↔ ChatML Dataset Converter Convert instruction-tuning data between Alpaca (instruction/input/output) and ChatML / OpenAI messages format, both directions, so the same dataset works across trainers.
• JSON Schema to GBNF Grammar Turn a JSON Schema into a GBNF grammar for llama.cpp and vLLM constrained decoding, so a local model is forced to emit valid structured JSON.
• Fine-Tuning JSONL Deduplicator De-duplicate a fine-tuning JSONL dataset — exact and normalized matching on the prompt/completion or messages — and report how many rows were removed.
• OpenAI logit_bias Builder Build an OpenAI logit_bias map from words or token IDs — ban or boost tokens — with a -100 to 100 slider, ready to paste into your API call.
• vLLM serve Command Generator Generate a vllm serve command from your model, GPU count, dtype, max-model-len and quantization — with tensor-parallel and the right flags filled in.
• MCP Tool Schema Validator Validate a Model Context Protocol tool definition — name, description and inputSchema — against the spec, and catch the JSON Schema mistakes that make a tool invisible to the model.
• LLM License & Commercial-Use Checker Check whether an open-weight model — Llama, Mistral, Qwen, DeepSeek, Gemma — is cleared for commercial use, and the catches (active-user caps, attribution, output restrictions).
• Assistants / Thread Cost Estimator Estimate the cost of an OpenAI Assistants or threaded conversation as message history grows — cumulative input tokens per turn quickly dominate the bill.
• pgvector Index Sizing Calculator Estimate the disk and RAM a pgvector index needs for N vectors at D dimensions — flat vs HNSW vs IVFFlat — so you can size Postgres before loading embeddings.
• Agent Loop Token Budget Estimator Estimate the token budget an agent loop burns — system prompt + tool schemas + growing scratchpad over K steps — and see where the multi-step cost comes from.
• LLM API Pricing Table Compare API pricing for GPT-4o, Claude, Gemini, Llama, DeepSeek, Mistral and more side by side — sort by input or output cost per million tokens and estimate your cost per call.
• Chinchilla Compute-Optimal Calculator Find the compute-optimal training tokens, FLOPs, and rough GPU cost for an LLM using the Chinchilla scaling laws — given a parameter count, a token budget, or a compute budget.
• Full Fine-Tuning VRAM Calculator Calculate the GPU memory a full fine-tune needs — weights, gradients, Adam optimizer states, and activations — and compare AdamW, 8-bit Adam, gradient checkpointing, and ZeRO sharding.
• Speculative Decoding Speedup Calculator Estimate the wall-clock speedup of speculative decoding from the draft acceptance rate, draft length, and draft-to-target cost ratio — and find the draft length that maximises throughput.