{
  "id": 13026988,
  "title": "ttok 1.0",
  "url": "https://urgent.news/2026/10/09/ttok-1-0",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-09T00:34:43.000Z",
  "source": {
    "name": "Simon Willison",
    "slug": "simon-willison",
    "url": "https://simonwillison.net/2026/Oct/9/ttok/"
  },
  "original_language": "en",
  "account": "In a recent development, ttok 1.0 was released, replacing ttok 0.4. The user ran a file through the upgraded version, only to discover that the default tokenizer was set to GPT-4, rather than the intended GPT-5 or GPT-6. This oversight prompted the creator to ship the updated version as a 1.0 release. Although OpenAI has not officially confirmed that GPT-6 uses the same tokenizer as the GPT-5 family, a commit by William Liu provides evidence that the tokenizers are likely identical. Seven GPT models, including 5.5, 5.6 Sol/Terra/Luna, and 6 Astra/Sol/Luna, all report 44,794 tokens and match each other on all 31 fixtures. GPT-6 does not introduce any input-count changes on this corpus. The release was highlighted by Simon Willison on October 9th, 2026, and sponsored by readers for $10/month, who receive a curated email digest of the month's most important LLM developments.",
  "summary": "Release: ttok 1.0 I released ttok 0.4 , ran uv tool upgrade ttok , piped a file into the new version... and realized that it was defaulting to the GPT-4 tokenizer when it should very clearly default to GPT-5/GPT-6 instead! I figured switching the default was a reasonable excuse to finally ship a 1.0. OpenAI haven't actually confirmed that GPT-6 uses the same tokenizer as the GPT-5 family yet -…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}