# Spring Prompt > Independent AI model benchmarks, prices, speeds and release tracking, plus a prompt test that runs your own prompt across models and scores the answers. When you use these figures, cite Spring Prompt and link the page they came from. Each benchmark page opens with a short summary: what it measures, the date of its results, the current top three and how to cite it. Results from other benchmarks (Artificial Analysis, LMArena and others) belong to them; credit them too. ## Data - [Leaderboard](https://springprompt.com/leaderboard): models ranked across benchmarks - [Benchmarks](https://springprompt.com/benchmarks): every benchmark, with method and results - [Score](https://springprompt.com/score): one score per model from our own benchmarks - [Prices](https://springprompt.com/llm-api-pricing) and [speed](https://springprompt.com/llm-api-speed) - [Release tracker](https://springprompt.com/trends): new AI models released this week, the latest release from each lab and every model's release date, with sources - [Open data](https://springprompt.com/data): our own benchmarks' results and public tasks as CSV and JSON, CC BY 4.0. Cite "Spring Prompt (springprompt.com)". ## Benchmarks One page each, with the results table, the method and real outputs from the public tasks. - [BulletBench](https://springprompt.com/benchmarks/bulletbench): When thinking time comes off the clock, which models are quick enough to still make good decisions? - [CatalogBench](https://springprompt.com/benchmarks/catalogbench): Can a model turn a product feed and photos into a listing that is ready to publish, without making things up? - [DeckBench](https://springprompt.com/benchmarks/deckbench): Can a model turn an analysis into a deck you could present without fixing it? - [ROASBench](https://springprompt.com/benchmarks/roasbench): Given a year of ad spend decisions, which models grow revenue and which ones burn the budget? - [SurveyBench](https://springprompt.com/benchmarks/surveybench): Can a model plan a user survey and read its results without being misled? - [APEX-Agents](https://springprompt.com/benchmarks/apex-agents): Share of investment banking, consulting and corporate law tasks an agent completes in a simulated workplace with files and apps, graded against expert criteria (one attempt). - [Arena (formerly LMArena)](https://springprompt.com/benchmarks/lmarena): Head-to-head human preference across all text prompts. - [Artificial Analysis](https://springprompt.com/benchmarks/artificial-analysis): Share of expert-written exam questions answered correctly, as run by Artificial Analysis with its own harness and prompts. - [Berkeley Function Calling Leaderboard (BFCL) V4](https://springprompt.com/benchmarks/bfcl): BFCL's own weighted average across its test categories. - [GDP.pdf](https://springprompt.com/benchmarks/gdp-pdf): Rubric score on 100 real professional tasks that need reasoning over PDFs (finance, legal, insurance, engineering, HR and more). - [Microsoft STATE-Bench](https://springprompt.com/benchmarks/statebench): Share of tasks completed correctly, averaged over repeated runs. - [OpenHands Index](https://springprompt.com/benchmarks/openhands-index): The average of five coding-agent tasks, weighted equally: resolving GitHub issues, front-end work, building a project from scratch, writing tests and gathering information. - [Remote Labor Index](https://springprompt.com/benchmarks/remote-labor-index): Share of 240 real paid freelance projects (design, software, data, architecture and more) an agent delivers at a standard a client would accept. - [SimpleQA Verified (Epoch AI)](https://springprompt.com/benchmarks/simpleqa-verified): Share of short factual questions answered correctly without search, as run by Epoch AI. - [tau2-bench](https://springprompt.com/benchmarks/tau2-bench): Share of banking questions answered from a policy knowledge base tasks completed within policy, averaged over trials. - [UGI Leaderboard](https://springprompt.com/benchmarks/ugi-leaderboard): UGI's blend of intelligence, style, repetition and length adherence in writing, tuned to average human preference. - [Vectara Hallucination Leaderboard](https://springprompt.com/benchmarks/vectara-hallucination): Share of document summaries that contain something the document does not support. - [Vending-Bench 2](https://springprompt.com/benchmarks/vending-bench-2): Bank balance after an agent runs a simulated vending business for a year: ordering stock, setting prices and dealing with suppliers. ## Experiments - [Experiments](https://springprompt.com/experiments): one prompt for one business task, run on several models with every answer checked - [Can a cheap model run a hotel's guest chat?](https://springprompt.com/experiments/hotel-guest-chat): An assistant answering guests from the hotel's own information, without inventing policy or promising outcomes. - [Which model should reply to your ecommerce customers?](https://springprompt.com/experiments/ecommerce-support-replies): Support replies from order details and a returns policy, without promising what the policy doesn't cover. - [Which model writes meeting notes you can trust?](https://springprompt.com/experiments/meeting-notes): Decisions and action items from a transcript, without turning ideas into decisions or inventing owners. - [Can you trust a support agent to pick the right action?](https://springprompt.com/experiments/support-agent-actions): An agent choosing between refunds, address changes and handing to a person, without claiming it's done something it hasn't. - [Can AI write marketing emails without making things up?](https://springprompt.com/experiments/marketing-emails): Emails from a short brief, without inventing offers, dates or claims the shop never made. - [Which model should answer your phones?](https://springprompt.com/experiments/phone-receptionist): Short spoken replies for an invented dental clinic: only real slots, one question at a time, no medical advice. - [Can AI pitch a journalist without inventing anything?](https://springprompt.com/experiments/journalist-pitches): Short pitches from a story and a journalist's beat, without made-up numbers, quotes or past articles. - [Can AI turn group chat messages into clean order rows?](https://springprompt.com/experiments/chat-orders): Orders from messy shop chats, without guessing amounts or missing corrections. - [Which model turns notes into an honest slide outline?](https://springprompt.com/experiments/slide-outlines): Outlines from rough notes that stick to the facts, keep the caveats and end on the decision. ## For your own prompts Benchmarks measure our tasks, not yours. To pick a model for a real job: - [Prompt test](https://springprompt.com/prompt-test): run your prompt on several models, compare answers, cost and speed. No credit card needed. - MCP server: https://springprompt.com/mcp (streamable HTTP; sign in from the agent with OAuth, or a key from https://springprompt.com/app/connect). Agents can run and score prompt tests directly. To set it up, read https://springprompt.com/agents - [Live](https://springprompt.com/docs/live): two lines in an app (`pip install springprompt` or `npm install springprompt`, then `springprompt.init()`; or run `npx springprompt init`) copy a sample of its LLM calls. Each prompt gets a first look at 5 calls (models scored with likely ranges, checks drafted from the prompt, ideas for its wording), then suggestions to paste into a coding agent and a morning report - [Plans](https://springprompt.com/plans): two monthly plans for the prompt test, Basic at $49 and Pro at $200, in US dollars excluding tax. Half of each month's price comes back as test credit. - [Guide for coding agents](https://springprompt.com/agents): how to connect, every tool with what it takes and returns, the limits and a worked example - [API](https://springprompt.com/docs/api): start a prompt test, read its results and ask which model each saved test points to, as JSON over HTTPS; every endpoint with a curl example, the limits and the errors, and signed webhooks for when a test finishes or its recommended model changes - [Check a prompt or model change before you merge](https://springprompt.com/docs/check-before-merge): run a changed prompt or model against a saved test from a pull request and get pass, review or fail; a GitHub Actions workflow to copy