{"id":83657,"topic":"ai","source":"东方财富","title":"讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线 - 东方财富","url":"https://emwap.eastmoney.com/a/202609163876060292.html","url_hash":"2a73ff3469f070c9cf8283b39b5c57080930df2a","author":"","summary":"<a href=\"https://news.google.com/rss/articles/CBMiY0FVX3lxTFB4WW93dU1Fc1oxSGR4WWRpMTZlcHp3Y1VBam1hYXZZdEJvY2xBdWtzcmVkY180dDg3d1FpOGM4d1FoamItN2d6SXVjcmQ0RkdudDNfQTdZYm5vYjhJQW10dkE1UQ?oc=5\" target=\"_blank\">讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线</a>&nbsp;&nbsp;<font color=\"#6f6f6f\">东方财富</font>","content":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）\n9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）","image_url":"https://g1.dfcfw.com/g4/202112/20211206191716.png","lang":"zh","published_at":"2026-09-16T08:23:41+00:00","fetched_at":"2026-09-16T09:15:08+00:00","status":"read","starred":0,"extract_state":"ok","summary_auto":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分…","cluster_id":3547235,"extract_retries":0,"extract_error":null,"contract_version":"news_item.v1","format_contract_version":"news_item_formats.v1","dedup_url":"https://emwap.eastmoney.com/a/202609163876060292.html","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 493 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":493,"summary_length":221,"usable_text_length":493,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":493,"summary_length":221}},"news_item":{"id":83657,"canonical_url":"https://emwap.eastmoney.com/a/202609163876060292.html","source_url":"https://emwap.eastmoney.com/a/202609163876060292.html","title":"讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线 - 东方财富","source_name":"东方财富","author":null,"published_at":"2026-09-16T08:23:41+00:00","locale":"zh","topic":"ai","tags":[],"rss_summary":"<a href=\"https://news.google.com/rss/articles/CBMiY0FVX3lxTFB4WW93dU1Fc1oxSGR4WWRpMTZlcHp3Y1VBam1hYXZZdEJvY2xBdWtzcmVkY180dDg3d1FpOGM4d1FoamItN2d6SXVjcmQ0RkdudDNfQTdZYm5vYjhJQW10dkE1UQ?oc=5\" target=\"_blank\">讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线</a>&nbsp;&nbsp;<font color=\"#6f6f6f\">东方财富</font>","full_text":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）\n9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）","excerpt":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分…","extraction":{"state":"ok","confidence":0.9,"error":null,"explanation":"High confidence: full text extraction produced 493 characters.","diagnostics_url":"/api/diagnose?url=https%3A//emwap.eastmoney.com/a/202609163876060292.html","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 493 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":493,"summary_length":221,"usable_text_length":493,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":493,"summary_length":221}}},"display_formats":["compact","card","full","digest_section","json"]},"daily_stack_record":{"title":"讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线 - 东方财富","url":"https://emwap.eastmoney.com/a/202609163876060292.html","summary":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分…","source":"东方财富","date":"2026-09-16T08:23:41+00:00","content":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）\n9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）","confidence":0.9,"diagnostics_url":"/api/diagnose?url=https%3A//emwap.eastmoney.com/a/202609163876060292.html","quality_bucket":"high","failure_kind":"none","retryable":false,"quality_reason":"High confidence: full text extraction produced 493 characters.","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 493 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":493,"summary_length":221,"usable_text_length":493,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":493,"summary_length":221}},"tags":[]},"fallback_formats":["markdown","json","html"],"actions":{"read":"/item/83657","export_markdown":"/api/items/83657/export?format=markdown","export_json":"/api/items/83657/export?format=json","diagnose":"/api/diagnose?url=https%3A//emwap.eastmoney.com/a/202609163876060292.html"},"formats":{"full":{"id":83657,"title":"讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线 - 东方财富","url":"https://emwap.eastmoney.com/a/202609163876060292.html","source":"东方财富","author":null,"published_at":"2026-09-16T08:23:41+00:00","locale":"zh","topic":"ai","tags":[],"excerpt":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分…","full_text":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）\n9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）","reading_time_min":1,"extraction":{"state":"ok","confidence":0.9,"error":null,"explanation":"High confidence: full text extraction produced 493 characters.","diagnostics_url":"/api/diagnose?url=https%3A//emwap.eastmoney.com/a/202609163876060292.html","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 493 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":493,"summary_length":221,"usable_text_length":493,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":493,"summary_length":221}}},"quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 493 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":493,"summary_length":221,"usable_text_length":493,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":493,"summary_length":221}},"actions":{"read":"/item/83657","export_markdown":"/api/items/83657/export?format=markdown","export_json":"/api/items/83657/export?format=json","diagnose":"/api/diagnose?url=https%3A//emwap.eastmoney.com/a/202609163876060292.html"}},"digest":{"id":83657,"title":"讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线 - 东方财富","url":"https://emwap.eastmoney.com/a/202609163876060292.html","source":"东方财富","topic":"ai","published_at":"2026-09-16T08:23:41+00:00","excerpt":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。 据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分…","quality_bucket":"high","quality_reason":"High confidence: full text extraction produced 493 characters.","reading_time_min":1,"cluster_id":3547235},"card":{"display_title":"讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线 - 东方财富","subtitle":"东方财富 · 2026-09-16","summary":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。…","badges":["quality:high"],"links":{"read":"/item/83657","original":"https://emwap.eastmoney.com/a/202609163876060292.html","diagnose":"/api/diagnose?url=https%3A//emwap.eastmoney.com/a/202609163876060292.html"},"quality_warning":null},"export":{"title":"讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线 - 东方财富","url":"https://emwap.eastmoney.com/a/202609163876060292.html","summary":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分…","source":"东方财富","date":"2026-09-16T08:23:41+00:00","content":"9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）\n9月16日，科大讯飞宣布讯飞星火全国产语音基座大模型Spark-Audio-1.0-Preview上线。\n据介绍，此次首发的Spark-Audio-1.0-Preview采用0.65B（Dense结构）的音频编码器，搭配30B-A3B（MoE结构）的大语言模型，使用了1300万小时音频以及大量文本数据，基于纯国产算力集群训练完成；在同一个模型中可输入文本和语音两个模态，支持语音转写、多语言翻译、多方言识别、环境音识别、说话人识别、情感分析以及复杂的音频问答等任务。\n（文章来源：第一财经）","confidence":0.9,"diagnostics_url":"/api/diagnose?url=https%3A//emwap.eastmoney.com/a/202609163876060292.html","quality_bucket":"high","failure_kind":"none","retryable":false,"quality_reason":"High confidence: full text extraction produced 493 characters.","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 493 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":493,"summary_length":221,"usable_text_length":493,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":493,"summary_length":221}},"tags":[],"format_contract_version":"news_item_formats.v1"}}}