Anthropic has drawn a line around its free Claude offering, applying a rolling five‑hour usage window that determines how many prompts a user can send before the system throttles access. The window begins with a user’s first prompt and does not reset at midnight, preventing people from gaming the limit by clustering requests at the end of the day. In practice, most users see the ability to send roughly 15 to 40 messages during each five‑hour span, though the exact number can fluctuate based on demand, prompt complexity and attachment size.
Behind every prompt lies a process called inference, where the model consumes tokens—numeric representations of words, characters and punctuation—to generate a response. Because each token costs compute resources, longer or more intricate questions drain a user’s allowance faster. Anthropic advises users to keep prompts concise, noting that higher‑effort settings, which produce more thorough answers, also chew through limits more quickly.
Free‑tier features and model access
Free accounts are restricted to two of Claude’s three models: Sonnet 4.6, which serves as the default, and Haiku 4.5. The flagship Opus 4.8 model, along with the newer Claude Code and Claude Design agents, remain exclusive to paid plans such as Pro, Max, Team and Enterprise. Despite these restrictions, the free tier still provides a suite of useful capabilities. Users can invoke web search, upload up to 20 files per chat (each file capped at 500 MB), and organize conversations with Projects and Artifacts—a feature that lets Claude generate small apps, games or interactive flashcards.
The platform also enforces a hard token ceiling of 200,000 tokens across all models and paid plans, a limit that applies regardless of subscription level. While the free tier does not include ads—a promise Anthropic reaffirmed after OpenAI’s sponsored‑response testing—it does require users to opt out of model‑training data collection if they prefer privacy. This setting resides in the Privacy section of the app’s menu and can be toggled off at any time.
Users seeking deeper coding assistance will find Claude Code unavailable on the free plan. The tool, which can generate, debug and explain code snippets, is a major draw for developers and remains locked behind paid subscriptions. Similarly, Claude Design, Anthropic’s design‑focused agent, is offered only as a preview to higher‑tier customers. Nevertheless, free users can still ask Claude coding‑related questions and copy‑paste the resulting snippets for one‑off help.
Anthropic’s rate‑limit policy also mentions that the number of messages allowed may vary with overall system demand, and the company reserves the right to impose additional limits to ensure fair access. Users have reported hitting the five‑hour cap after a single complex Claude Code prompt, underscoring how token‑heavy interactions can accelerate throttling.
Overall, the free tier delivers a functional, ad‑free chatbot experience with access to the company’s most capable model outside the flagship offering. By balancing token consumption, model availability and feature set, Anthropic aims to keep casual users engaged while nudging power users toward paid subscriptions for higher‑effort processing, advanced agents and fewer usage constraints.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory SEO vous aide à automatiser le contenu d'actualités pour votre site.