Maritaca S5-Free API: 1M Context for $0 — While the Launch Offer Lasts (2026)
Brazilian AI lab Maritaca released S5, the newest generation of its Sabiá model family, on November 3 — and alongside it, a free API variant called S5-free with no token charges. The headline numbers are unusually generous for a free tier: a 1 million-token context window and 60 requests a minute. The catch is in the fine print: S5-free is a limited-time launch offer that can end at any time, and unlike the no-sign-up tools in our directory, it requires a Maritaca platform account. Here is exactly what it gives you, what it doesn't, and when the paid fallback kicks in.
What S5-free actually is
S5 is Maritaca's best model to date, built for complex analysis of hundreds of documents, deep web research, and code-agent work — Maritaca's own launch post shows it driving Aider and OpenClaw sessions, including building a small Python project and running its test suite. The free variant, S5-free, is the same model served through the API with no token charges. It offers adjustable reasoning effort and a 1 million-token context window, and it is strongest in Portuguese and Brazilian topics, though it also handles coding and long documents in English. One honest framing from the launch coverage: this is a launch promotion, not a permanent free tier. Maritaca can end it whenever it chooses, at which point the same calls would bill at the paid S5 rate.
That distinction matters for anyone building on it. A standing free tier you can plan around is one thing; a launch offer is a window to try the model for real work, not a foundation to build a product on. Our free APIs directory lists the standing tiers separately for exactly this reason.
The limits, in plain numbers
Every figure below comes from Maritaca's pricing and rate-limits documentation as reported on November 6, 2026. Free tiers change often, so check the docs before you build on these numbers.
60 requests per minute. The same rate limit applies to every account, free or paid. For a single developer testing an agent loop, that is plenty; for anything serving many users at once, it is the first wall you will hit.
256K tokens per minute, in and out. Input and output each get a 256K-token-per-minute allowance. With a 1M-token context window, that means you cannot actually fill the whole window in a single minute of calls — the rate limit, not the window, sets the practical ceiling on how much you can push through at once.
Text input only. S5-free accepts text. It does not include the Batch API, Flex, or the built-in web search and code execution tools that the paid model offers. If your agent needs the model to search the web or run code itself, this free variant will not do it — you would need to bring your own tools, which is exactly how Maritaca's own Aider and OpenClaw demos use it.
An account is required. You sign up on the Maritaca platform and create an API key. This is a free API, not a no-sign-up tool — if you want something that works with no account at all, start with the verified no-sign-up agents in our directory instead.
How to call it
S5-free speaks the OpenAI-compatible API, so any client that accepts a custom base URL works. Point it at https://chat.maritaca.ai/api with the model name S5-free and your Maritaca API key. Maritaca's launch post demonstrates it with coding agents that accept a custom API — Aider for a terminal workflow and OpenClaw for an autonomous agent — which makes it a natural fit for the coding agents that run without their own API key pattern: the agent is free, and the model behind it is free too, at least while the offer lasts.
One practical tip before you wire it into anything: estimate the bill for the day the offer ends. Agent Toolbox is a free, no-sign-up set of browser tools for agent builders — its token and cost estimator shows per-call and monthly cost across major models side by side — and it added an MCP server this week. Knowing what S5 would cost at the paid rate tells you whether your workflow survives the end of the free window.
What happens when the free offer ends
The fallback is the paid S5 model, priced at R$0.50 per million input tokens and R$2.50 per million output tokens according to Maritaca's launch post — the lowest price in the Sabiá family. Reasoning tokens are billed at the output rate. There is no published end date for the free variant, so the honest way to treat S5-free is as a generous trial: use it to find out whether S5 is good enough for your workload, keep your prompts and client configuration portable, and be ready to either pay the (low) per-token price or move to another free tier when the window closes.
For scale, compare the shape of the offer with the free tiers we track elsewhere. OpenRouter's free model variants are capped at 50 requests a day for accounts that have bought fewer than 10 credits, rising to 1,000 a day after a one-time $10 top-up, according to the daily-verified tracker updated November 4. S5-free's 60 requests a minute is a far higher ceiling — but it is one model, text-only, on a launch offer, where OpenRouter's free tier is standing. Generous and temporary beats stingy and permanent for testing; it loses for anything you need to still be running in six months.
Who should try it, and who should skip it
Try it if you build coding agents or document-analysis workflows, you work in Portuguese or with Brazilian content, or you want a genuinely large context window to test long-document tasks without paying. The 1M-token window is the real draw — few free tiers offer anything near it.
Skip it if you need a no-account tool, image or multimodal input, built-in web search, or a free tier you can rely on long term. In those cases the standing free tiers in our free APIs directory and the best free coding agents we tested are the safer starting point.
Ground rules
Don't build a product on a launch offer. Keep your code provider-agnostic — an OpenAI-compatible base URL is easy to swap — so the end of S5-free costs you a configuration change, not a rewrite.
Verify the limits yourself. Free tiers change without notice. Re-check Maritaca's pricing and rate-limits pages before you rely on the 60-requests-a-minute figure for anything that matters.
Never paste secrets. No passwords, no API keys, no unreleased financials. S5-free needs an account, which means your prompts travel to Maritaca's servers — treat it like any cloud API, not like a local model.
FAQ
Is Maritaca S5-free really free?
Yes, for now: the S5-free model carries no token charges. But it is a limited-time launch offer with no published end date, not a permanent free tier. The paid S5 model costs R$0.50 per million input tokens and R$2.50 per million output tokens.
Do I need an account to use S5-free?
Yes. You need a Maritaca platform account and an API key. It is a free API, not a no-sign-up tool.
What are the rate limits?
60 requests per minute, and 256K tokens per minute each for input and output, the same for every account. The model accepts text input only.
What is the context window?
1 million tokens, with adjustable reasoning effort. S5 is strongest in Portuguese and Brazilian topics and also handles coding and long documents.
Can I use it with my coding agent?
Yes — it is OpenAI-compatible, so tools that accept a custom API base URL work. Maritaca's launch post demonstrates it with Aider and OpenClaw. Set the base URL to https://chat.maritaca.ai/api and the model to S5-free.
What should I use when the offer ends?
Either pay the S5 per-token rate or move to a standing free tier. Our free APIs directory tracks the ones that don't expire.