LanguageModel (Prompt API)checking…
Multimodal Gemini Nano. Inputs: text, image, audio (audio needs GPU).
Stable in extensions since Chrome 138; web origin trial Chrome 148.
Sampling params (temperature/topK) work in any extension and on the web from Chrome 148+; on older web Chrome they're silently ignored.
Languages: en, ja, es.
⚙️ Options (temp, topK, system prompt, JSON schema, probe languages)
JSON response schema (when "JSON-constrained" is on)
tokens: 0 / 0
0%
Summarizernot checked
Condenses long-form content. Stable in Chrome 138+. type:
tldr/key-points/teaser/headline. format: markdown/plain-text.
length: short/medium/long. No documented input cap, but very long inputs may be truncated.
Translatornot checked
Pairwise translator (one source → target per instance). Chrome 138+.
Each language pair downloads its own model on first use. Translations are processed sequentially per instance.
Supported: ar, bg, bn, cs, da, de, el, en, es, fi, fr, hi, hr, hu, id, it, iw, ja, kn, ko, lt, mr, nl, no, pl, pt, ro, ru, sk, sl, sv, ta, te, th, tr, uk, vi, zh, zh-Hant.
LanguageDetectornot checked
Returns ranked { detectedLanguage, confidence } array. Chrome 138+.
Accuracy is poor for very short input (single words / fragments).
Writernot checked
Composes new text from a brief. Origin trial (Chrome 137–148+) — register at
developer.chrome.com/origintrials if running on a normal page, or enable the
matching chrome://flags. Inputs feed a single one-shot generation; for multi-turn use the Prompt API.
Rewriternot checked
Rewrites existing text — adjusts tone, length, formality. Origin trial alongside Writer.
Useful for "make this more formal", "shorten", etc.
Proofreadernot checked
Returns { correctedInput, corrections[] } with start/end indices and per-error type
(grammar/spelling/punctuation). Origin trial only (Chrome 141–145).
May be unavailable depending on your channel — Canary/Dev are most reliable.