Token
The basic unit of data a language model works with. Before the model gets a text, a separate program splits it into tokens. A common English word usually becomes a single token. Rarer words are broken into several pieces, regardless of their meaning. Images and sound are turned into tokens as well. The limit on how much a model handles at once is counted in tokens. Where use is billed by volume, the price is also set per token. Languages that are rare in a model's training data need more tokens to say the same thing. Working in them therefore costs more and reaches that limit sooner.
AI underdogs
…can also be the selling point without any openness. It charges $2 for every million tokens the model reads and $6 for every million it writes. xAI sells it inside coding tools such as Cursor and GitHub Copilot.9 Grok is a name many people know, yet Grok 4.7 sits…
AI doesn’t think. AI computes.
…a term from computational linguistics, not the humanities: a statistical sequence of tokens, with no claim to comprehension, reasoning, or anything the older ideas of language would promise. Same goes for "model", which is a representation of a thing by…
AI runs on borrowed money
…the expensive top end, that is easy to miss. . DeepSeek charges up to $4 per million tokens of generated text.5 Anthropic charges $20 for the same amount from Opus 5.5.6 Their ability trails the top end. On the Artificial Analysis index, Opus 5.5 scores 58 and the…