{"rewrite":{"id":"r_48245da10466938bf7419c8d","clusterId":"c_855ea6d75cfb9e3cfa65f6cf","slug":"third-intelligence-releases-haas-ja-v1-0-japanese-speech-recognition-model","model":"deepseek-v4-1-flash","headline":"Third Intelligence Releases Haas-ja-v1.0 Japanese Speech Recognition Model","summary":"Third Intelligence released Haas-ja-v1.0, a Japanese-specialized speech recognition model with 615.8 million parameters, as open source. The company says it has the top accuracy among the open-weight models it evaluated, scoring above the commercial model Gemini 3.5 transcribe. A free demo transcribed 4 minutes 19 seconds of audio in 0.32 seconds after upload, with some errors that needed manual correction. The model is on Hugging Face under Apache License 2.0, with inference code on GitHub.","whyItMatters":"A 615.8 million parameter Japanese ASR model that outscores a commercial model on the company's own accuracy chart and runs at low cost puts a lightweight open option in front of teams that would otherwise pay API fees.","webCardHtml":"\u003cp\u003eThird Intelligence published the model weights on Hugging Face and the inference code on GitHub under Apache License 2.0. The company\u0026#39;s cost chart compares commercial models by API fees and open models by the fee for inference on an H200 in the Japan region, and places Haas-ja-v1.0 as the lowest-cost model at equivalent accuracy.\u003c/p\u003e\u003cp\u003eIn a hands-on test, a 4 minute 19 second clip finished transcribing 0.32 seconds after upload. The output still needed manual fixes: a name rendered as a similar-sounding wrong name, and a title likewise. Speed and accuracy claims come from Third Intelligence\u0026#39;s own evaluation, not an independent benchmark.\u003c/p\u003e","blueskyPost":"Haas-ja-v1.0 runs under Apache License 2.0 on Hugging Face, so anyone can fine-tune it. The demo's manual corrections suggest the accuracy claim holds on clean audio, not on messy real-world recordings.","twitterPost":"Haas-ja-v1.0 is Apache 2.0 on Hugging Face, so rivals can build on it. The demo's manual errors hint the accuracy ceiling is clean audio.","threadsPost":"Haas-ja-v1.0 ships under Apache License 2.0 on Hugging Face, which means the weights are open to fine-tuning by anyone. The free demo still needed manual corrections, so Third Intelligence's top-accuracy claim rests on clean audio, not the messy recordings most transcription work involves.","newsletterBlurb":"Third Intelligence has released Haas-ja-v1.0, a Japanese-specialized speech recognition model with 615.8 million parameters, as open source under Apache License 2.0. The company says it scores highest among the open-weight models it evaluated, above the commercial Gemini 3.5 transcribe. A free demo transcribed a 4 minute 19 second clip in 0.32 seconds, though the output needed manual correction in places.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20261102-haas-ja-v1-asr/\",\"title\":\"Japanese-specialized, high-speed and highly accurate transcription AI \\\"Haas-ja-v1.0\\\" debuts, recording the highest performance among open models\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":5618,"outputTokens":832,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1793603661,"createdAt":"2026-11-02T07:11:09.000Z","publishedAt":"2026-11-02T07:12:59.000Z","updatedAt":"2026-11-02T07:12:59.000Z"},"cluster":{"id":"c_855ea6d75cfb9e3cfa65f6cf","canonicalTitle":"日本語特化で高速かつ高精度な文字起こしAI「Haas-ja-v1.0」が登場、オープンモデルの中で最高性能を記録","representativeArticleId":"a_9b33588729d3d9d9e3efec5b","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"Haas-ja-v1.0\"],\"studios\":[\"Third Intelligence\"],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-11-02T06:24:00.000Z","lastSeenAt":"2026-11-02T06:24:00.000Z","updatedAt":"2026-11-02T07:13:00.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20261102-haas-ja-v1-asr/","title":"日本語特化で高速かつ高精度な文字起こしAI「Haas-ja-v1.0」が登場、オープンモデルの中で最高性能を記録"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["Haas-ja-v1.0"],"studios":["Third Intelligence"],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":["Third Intelligence released Haas-ja-v1.0, a Japanese-specialized speech recognition model with 615.8 million parameters.","The company says Haas-ja-v1.0 scored highest among the open-weight models it evaluated, above the commercial model Gemini 3.5 transcribe.","A free demo transcribed 4 minutes 19 seconds of audio in 0.32 seconds after upload, with some errors requiring manual correction.","Haas-ja-v1.0 is available on Hugging Face under Apache License 2.0, with inference code on GitHub."],"pageLinks":{"healthSlugs":["haas-ja-v1-0","third-intelligence"],"storySlugs":[],"entities":[{"slug":"haas-ja-v1-0","supported":0,"disputed":0,"contradicted":0,"unverified":3,"overall":"unverified","story_count":1,"has_dossier":false},{"slug":"third-intelligence","supported":0,"disputed":0,"contradicted":0,"unverified":1,"overall":"unverified","story_count":1,"has_dossier":false}],"stories":[],"entitiesOk":true,"storiesOk":true}}