Is Contexto free? And thirteen more questions, answered

Contexto is a daily word game in which one word is hidden and every guess you make comes back with a rank — its position in a list of about 30,000 words sorted by how close each one sits in meaning to the answer. Rank 1 is the answer. There is no letter feedback, no guess limit and no way to lose. One new word opens every midnight, and everyone playing in your language gets the same one. Our complete guide explains the ranking in detail, or you can just play today's puzzle and work it out from the numbers.

Yes. The web game is free, needs no account and has no paywall — your streak and history are kept in your browser's local storage rather than on a server. It is funded by advertising. The official mobile apps offer an optional paid tier that removes ads and unlocks the full back catalogue, but nothing about the daily puzzle is behind it.

Contexto was built by Nildo Junior, a Brazilian developer based in Norway, and launched in 2022. It began as a Portuguese-language project — the first audience was Brazilian — and the English and Spanish editions followed once it took off. It is one of the few games in this genre that was not an English game translated outward.

Yes, for both platforms. Contexto: Word Guess on the App Store and Contexto on Google Play are the official releases; Android users who cannot reach Google Play can install the APK from APKPure. There is also a Contexto on Stream Chrome extension for streamers who want the board on screen. The apps play the same daily puzzle as the browser version and add an offline archive and an ad-free option. Be careful in the app stores: the game's popularity has attracted clones with near-identical names and icons.

A language model reads a very large amount of text and records, for every word, the contexts that word appears in — which words show up nearby, in what kinds of sentences, how often. That record is compressed into a vector of a few hundred numbers. Words used in similar ways end up with similar vectors. Contexto measures the distance from the answer's vector to every other word's, sorts the whole dictionary by that distance, and your guess's rank is its line number in that sorted list. The consequence worth remembering is that the model has never seen a dictionary: it reflects usage, not definitions. The full guide works through what that means in practice.

Green covers ranks 1 to 300 — the answer's own neighbourhood, and the signal that you are in the right conceptual region. Yellow and orange cover 301 to 1,500, meaning related by category rather than by closeness. Red is 1,501 and above, which in practice means unrelated; the exact number in the red band carries almost no usable information. The rank is always shown next to the colour, so nothing in the game depends on being able to tell the bands apart visually.

When you ask for a tip, Contexto gives you a word rather than the answer, and the difficulty setting controls how helpful that word is. On Easy, tips come from very close to the answer — often inside the top 20 — and usually end the round quickly. On Medium they land in the middle of the green band, around ranks 50 to 300, which relocates you without finishing the job. On Hard they come from the yellow band and tell you only the general category. Most regular players settle on Medium.

No. Guesses are reduced to a base form before they are ranked, so "running" is scored as "run" and "moons" as "moon". You cannot extract extra information by cycling through the forms of a word. Spelling, on the other hand, is never corrected — a typo is ranked as whatever string you actually typed, which usually lands past 20,000 and can look misleadingly like a real signal.

Under 15 is an excellent round and usually means an early guess got lucky. Between 15 and 40 is a good, normal round, and where most experienced players live. Between 40 and 80 is ordinary, or a hard word played well. Over 100 usually means committing to the wrong category early and not testing out of it. The average across a typical week is around 40, but it swings hard by word type: concrete nouns like anchor or compass average under 30, while abstract nouns like rumour run much higher with a far wider spread. The strategy guide is aimed squarely at the "over 100" days.

Yes. Every previous puzzle stays available, and past games do not affect your streak, which makes them the best way to practise — you can reveal the answer immediately and check your reasoning against the real ranking. Our archive lists every puzzle we have published with its answer, the ten closest words and a note on how the day went.

Yes, and they are genuinely separate games rather than translations. Each language runs its own word list and its own model, so the answer on any given day is different in each one — you can play all three without spoiling any of them. The Portuguese answer page and the Spanish one are updated daily alongside the English. Words that only really exist in one language, like saudade or sobremesa, make excellent puzzles in that language and would be impossible in the others.

Both hide a word and score guesses by meaning, but they report it differently and that changes everything. Contexto gives you a rank out of about 30,000, so every guess — even a terrible one — carries usable information. Semantle gives you a raw cosine-similarity score, and because almost every English word scores between 0 and 8 against any given target, most guesses come back indistinguishable from each other. Contexto also has colour bands, a built-in tip system and a local-midnight reset; Semantle has none of those and resets at midnight UTC. Semantle is substantially harder, and there is a full comparison on our Semantle page.

Because the model measures how words are used together, not what they mean. Pairs that feel connected to a person can share very little actual context in written text — "sun" and "moon" are the classic example, where one can sit at rank 9 and the other past 200. The reverse also happens: antonyms like "hot" and "cold" appear in nearly identical sentences, so the model puts them very close together. Learning where your intuition and the corpus disagree is most of the skill in the game.

The usual causes, in order of likelihood: an ad blocker or privacy extension blocking the script that fetches the day's puzzle; a browser set to clear local storage, which wipes your progress and can leave the board in a confused state; or a stale service worker from an older version. Try a hard refresh first, then a private window with extensions disabled. If the board loads but your streak has vanished, that is local storage having been cleared — it is stored only in your browser, so it cannot be recovered from a server.

No. There is no guess limit and no fail state — a bad round just gets long. That is a deliberate design choice: the game gives you a tip system and a give-up button instead of a loss, so the worst outcome is that you learn the word rather than that you lose the day. If you do give up, it is worth reading the closest words afterwards, because seeing the neighbourhood you failed to find is what makes you recognise it next time.