Claude Haiku 5.5, released October 7, 2026, is Anthropic’s newest Haiku model, available on Free, Pro, Max, Team and Enterprise plans. Its API list price is 90% below Haiku 4.5’s for prompts up to 100,000 tokens, and it adds an effort setting. Anthropic’s system card says it hallucinated more than other recent models in targeted honesty tests, about as much as Haiku 4.5. This Tier C Starter Guide, built from Anthropic’s pages and Artificial Analysis’s results, shows when to use it, when to consider Sonnet 5.5 and how to check a first result. FSR didn’t run Haiku 5.5 for this guide.
ChatGPT dots are agents OpenAI calls always-on. Your first dot is included in Pro or Business Premium at no extra cost, OpenAI says. A dot can use plugins you’ve already connected: in FSR’s test, a new dot showed two suggestions drawn from email before FSR had sent it a message. This Tier B Starter Guide, built from OpenAI’s public pages and FSR’s tests of one dot over four days, shows who can start, three first jobs with templates, and what Pause did. FSR didn’t test a dot with the computer off, and nothing it tested is, on its own, a reason to upgrade a plan for a dot.
Leonardo.Ai is an AI image and video platform. Its Free plan gives 150 tokens a day: usage credits, not images. Its Privacy Policy says what you generate on Free is public by default, and Help says private generation and deleting generated images need a paid plan. Essential, the lowest-priced paid Solo plan, is listed at $12 a month on monthly billing, before tax. This Tier C Starter Guide, built from Leonardo.Ai’s public pages and a Canva Help page, shows when to test on Free, pay, keep your current tool or wait. FSR didn’t use Leonardo.Ai for this guide.
MiMo-V2.6 is Xiaomi’s AI model family, with five ways in: the Desktop app, the pay-as-you-go API, Token Plan, a third-party host and the open weights. Xiaomi lists API input for its Pro model at $0.0036 per 1M tokens for cache hits only; a cache miss is $0.435. Token Plan is for coding tools only and can’t be refunded, and the Desktop app is licensed for “non-commercial purposes,” which its agreement doesn’t define. This Tier C Starter Guide, built from public documents, shows what each route requires and what to check before you pay. FSR used none of the five for this guide.
Base44 is an AI app builder: describe an app in plain language and it generates one. Its docs say a Free workspace can build public apps only. Private apps start at Starter, listed at $20 a month on monthly billing. On every plan except Enterprise, workspace data can be used to train AI models, with no opt-out. This Tier C Starter Guide, built from Base44’s public pages, shows what to build first and what to check before you pay or share. FSR built and tested nothing for this guide.
ChatGPT Pro 200 is still $200 a month, with a lower usage allowance. At FSR’s last read, OpenAI’s help page didn’t say how much lower. If your Pro 200 subscription was active at any point from September 22 through 10 a.m. Pacific Time on September 29, 2026, you keep the previous allowance through October 29 while you stay on Pro 200. Plan around your billing date, not October 29. This Tier C Starter Guide draws on OpenAI’s public pages, a press report and an X post. FSR ran no tests.
OpenAI’s DevDay 2026 recap describes more than 20 major announcements, from always-on agents called dots to a GPT-6.1 Sol model, ChatGPT Space and a $500 Pro plan. Many depend on your plan, region, age, device or admin settings, and being eligible doesn’t mean a feature has reached your account. This Tier C Starter Guide, built from OpenAI’s public pages, shows who each one is for, how to check one against work you already do, and when to wait. FSR tested none of them.
This Tier C Starter Guide, built from the providers’ public pages, helps you match one first job to one of eight AI tools: ChatGPT, Claude, Gemini, Grok, Perplexity, Runable, Kimi or Mistral Vibe. FSR tested none of them. It names no default tool, because free use comes with conditions: several tools use chats for training unless you opt out, share links can be public and credits run out. It shows what to check before you type, how to test once with invented material and when to stop.
Claude Sonnet 5.5 keeps Sonnet 5’s API price, and Anthropic says it costs up to 30% less per task than Sonnet 5. Independent evaluator Artificial Analysis found a pre-release build cheaper per task than Sonnet 5 and Opus 5.5 at Medium effort but costlier at Max, scoring between the two at both levels. This Tier C Starter Guide, built from those public sources, explains who can use it, which model to try first, and what to check.
What Runable is, what its Free plan covers, where its own pages disagree, and a safe first test built from invented material, with a stop before any payment.
