Home / AI Chat Models
AI chat models, tested on free accounts
How large language models actually serve a review blog: what the free tier really gives you, which model writes a usable product review, and the prompts that get there. Every number below came from a free account on our own screen.
Model comparisons
Four head-to-head tests. Identical prompts, identical files, free tier against free tier — we score what came back, not what the spec sheet promises.
DeepSeek vs ChatGPT: We Gave Both Free Versions the Same 6 Questions
The same 6 questions on free DeepSeek and ChatGPT — email, math, code, translation — then 10 rapid-fire messages to find the real limit.
DeepSeek vs ChatGPT vs Gemini in 2026: The Same 4 Free Tests — and One Answer We Did Not Expect
Four identical tests on free DeepSeek, ChatGPT and Gemini: writing, math with a known answer, a resume PDF, and a Chinese round.
Gemini vs ChatGPT: We Gave Both Free Versions the Same 4 Tests
A logic puzzle, a strict writing brief, a fresh-news question and a messy table — run on free Gemini and ChatGPT. Full scorecard.
ChatGPT vs Claude: We Ran the Same 3 Free Tests — Claude Never Replied Once
Same 3 tests on ChatGPT free and Claude free: ChatGPT answered everything; Claude locked replies behind an ID verification wall.
Free tiers & pricing
What $0 buys on each model: message caps, request windows, context limits and every plan price, read from our own accounts and the vendors' own pages.
See all 11 free-tier limits and prices →
ChatGPT Free Plan Limits in 2026
The 10-messages-every-5-hours figure is out of date. What the free plan actually limits now, what the notices say word for word, and which model $0 gets you.
GPT-6 in ChatGPT: What Free Users Actually Get (Tested on a Free Account)
GPT-6 became the free default on October 8. Six real prompts, timed and screenshotted: a calculator, a map, charts, forms — and limits.
Does DeepSeek Have a Free API Tier? What the Docs Say
DeepSeek's pricing page lists per-token rates, not a free monthly quota. What the documentation supports, what third-party pages claim, checked on 3 November 2026.
DeepSeek Pricing in 2026: The App Is Free, the API Is Not
No subscription exists. The chat app is free; only the API is billed. Exact rates, the off-peak half-price window, and the real cost of one conversation.
DeepSeek API Cost Calculator
A working calculator, the exact per-million rates behind it, and a worked comparison of which cost lever is worth pulling for your workload.
DeepSeek's 1M-Token Context Window: Limits and Costs
1M tokens is not a number most people can picture. Here it is in characters, words and paperbacks - plus the 384K output cap nobody reads correctly.
Grok Free Tier, Tested: 30 Requests per 24 Hours, Two Models, and the Exact Limit Screen
Free Grok gives you two models and 30 chat requests per rolling 24 hours. We measured it request by request on a real free account on 5 November 2026, including…
Grok Pricing: What $0, $8, $30, $100 and $200 Actually Include
Five individual plans from $0 to $200 plus Teams, first-month promos the pricing page does not show, API access billed separately, and the old SuperGrok plan…
Grok API Pricing: Every Model Rate, the 200k Rule, and What One Request Costs
grok-4.7 costs $2 per million input tokens and $6 per million output tokens; prompts of 200k tokens or more double the rate for the whole request. Full tables,…
How to Subscribe to Grok: Every Upgrade Path in the App, the Fine Print, and the 24-Hour Cancel Rule
Every way the Grok app asks you to upgrade, the plans and first-month promo prices, and the footer fine print: prices exclude tax, subscriptions auto-renew, and…
What Is SuperGrok? The Old Plan Name and the Four Current Tiers
The Grok app still gates Auto, Expert and Heavy with 'Requires SuperGrok' - a plan name the pricing page no longer sells. What the app's model picker and plans…
Beginner tutorials & fixes
From your first chat to API keys, 429s, the 402 error, verification walls and the dsh command line — the steps we walked, with screenshots.
See all 12 beginner guides and fixes →
DeepSeek for Beginners: Your First Chat in 10 Minutes
Beginner's DeepSeek guide with real screenshots: first message, DeepThink vs Search, file uploads, chat history, five copy-paste prompt starters and a fix-it table.
ChatGPT "Something Went Wrong": We Tried to Trigger It 6 Ways on a Free Account — Only One Error Showed Up
Zip uploads, a 20 MB file, a 250 KB paste, offline sending, three tabs at once — almost none of the famous ChatGPT errors happen anymore. The one that did,…
Claude's Verification Wall: What "Complete Verification to Continue" Actually Asks For
Three screens, one government ID, and a phone camera. The check we hit on a free Claude account, plus the rules Anthropic publishes and the group it is meant to…
DeepSeek API Key: Create One, Fix the Common Errors
The button is the easy part. Two model names, two base URLs, four error codes, and the blank lines that break JSON parsers.
DeepSeek Base URL: The Two Official Addresses, and Why Wrong Ones Still Return 401
Two base URLs, one per SDK format. We also sent unauthenticated requests: every tested path returned the same 401, including one that cannot exist. What that…
DeepSeek 429 Too Many Requests: Why Creating Another API Key Won't Fix It
A 429 from DeepSeek is a concurrency cap counted per account. The rate-limit page lists the two caps and no requests-per-minute figure - here is what to check…
DeepSeek 402 Insufficient Balance: Causes and How to Fix It
The error that looks like an authentication problem and is actually a wallet problem. Plus the balance endpoint that tells you which wallet is empty.
DeepThink Mode in DeepSeek: On vs Off, Tested on the Same Questions
DeepSeek asked the same riddles and trip-planning tasks with DeepThink on and off — every answer timed, screenshots included.
DeepSeek Harness CLI: What Each dsh Entry Point Does Before You Have an API Key
Tested: dsh headless exits 1 with MISSING_CREDENTIAL without a key and reports a masked key plus request_id with a bad one; dsh tui fails because no tui profile…
How to Install DeepSeek Harness (dsh) on Windows: What We Hit on Each Route
A tested Windows install guide for DeepSeek Harness: the npm CLI, the official desktop installer, the zod startup error we hit on a cold npx cache, and the fake…
DeepSeek Harness Plugins: Counting What 'Everything Is a Plugin' Means in Practice
Tested on 0.2.1-rc.5: the web profile composes 191 plugin entries, 32 disabled by default; dsh plugin forwards to pnpm and fails without it; adding an npm…
Grok Models Explained: What the App Menu, the Rate-Limit Buckets, and the API Rate Card Each Call Them
The Grok app menu (Fast, Build, Auto, Expert, Heavy) and the API model list (grok-4.7 at $2/$6 per million tokens) use different names with no official…
Prompts & agents for review writing
The sub-category this site is building toward: prompt sets and agent workflows that turn a raw free-tier model into a draft you can publish. First guides are in testing now.
The first guides in this sub-category are in testing now. They get published here once they pass the same method as everything else on the site — free account, real screenshots, dated numbers.
Looking for everything on the site in one list? The full guide index has all 45 guides, and the data page collects every figure we have measured.
How all of this was tested
Free accounts only, screenshots at every step, and a date on every number. When a test failed, the failure is published instead of an estimate. The full method — including the tests we could not finish — is on our how we test page. Affiliate links never change a verdict; when a page uses one, the disclosure says so.