GPT, Claude or Gemini? Every AI model side by side
You need to decide: which AI bot to plug into your project? ChatGPT from OpenAI, Claude from Anthropic, or Gemini from Google? Each company has several versions: cheaper and simpler, pricier and smarter, fast, *"for hard jobs"*. The internet has a thousand opinions on which is best. You need actual numbers.
This tool puts every current model in one table. You see: the rate for messages you send to the bot, the rate for the bot's replies, how much the bot fits in one chat, whether it understands images, and the model identifier (the code you paste into your app).
Filter by company, sort by price, tick *"must understand images"* and you get a short list in seconds. A decision in 30 seconds, based on official price lists, not someone's hot take.
How to use it
- Provider filter: leave *"all"* or pick one company (OpenAI / Anthropic / Google).
- Required features: tick *"understands images"* if the bot needs to read photos. Tick *"can use tools"* if you are building a bot that fetches data from your systems (search a DB, call an API).
- Sorting: click a column header: price per question, price per reply, how much fits in one chat, or alphabetical.
- Each row shows: standard rate, discounted rate for repeated chunks, how much the bot fits in one chat, how much it can write back, plus a short note "when to pick it".
- Model identifier is the code you paste into your app (for example gpt-4o or claude-sonnet-4-5). Copy it straight from the table to avoid typos.
When this is useful
Six typical situations where this comparator gives you a concrete answer instead of a guess:
- Picking a bot for a new project. You have 7 models on screen, one next to the other. Sort by price-per-reply (what you actually pay for what the bot writes). You will see the cheapest can be 100 times cheaper than the most expensive. The decision stops being a guess.
- Hunting for savings. You currently use GPT-4o, the bills are climbing. Sort ascending by price, you spot Gemini Flash, 30 times cheaper for the same workload. Run a test on a sample of your data, check quality, decide on facts.
- Checking whether a bot understands images. Tick the *"understands images"* filter and instantly see which models accept a photo of a receipt, a screenshot or a chart. No digging through each company's docs separately.
- Comparing the discount for repeated chunks. If your bot keeps sending the same long instructions every time (e.g. 5 pages of company rules), some providers give a discount: Claude drops the price by 90% (from 3 dollars to 30 cents per million words), OpenAI gives 50% off, Google has no such thing. The table shows it plainly, no asterisks and footnotes.
- Fitting a long document into the bot (a book, an hour-long meeting transcript, a stack of contracts). Sort by how much the bot fits in one chat: Gemini 2.5 Pro about 1.5 million words (roughly 6,000 pages), Claude about 500 pages, GPT-4o about 300 pages. For very long texts, the choice is obvious.
- Copying the model identifier into your code, typo-free. Click the identifier field in the table, copy, paste into your code. No more *"model not found"* errors.