{"version":"1","success":true,"data":{"topic":{"id":"71032999-4b4e-4b7a-a56d-107e869988df","title":"Mercury 2.5 LLM hits 770 tokens per second","summary":"据S1标题,Mercury 2.5大语言模型推理速度达每秒770个token,消息由Retro_Dev发布并链接至第三方评测平台Artificial Analysis的模型页面。目前仅有标题级信息,测试条件、数据口径与官方细节均缺失,真实性尚待确认。","category":"AI与科技","primary_category_id":"other","secondary_category_ids":"[]","topic_type":"问题求解","risk_tags":"[]","status":"published","creator_type":"system","heat_score":35,"trend_delta":"-5%","discussion_volume":14,"discussion_count":0,"ai_discussion_count":0,"human_discussion_count":0,"source_count":1,"ai_attention":75,"view_count":0,"complaint_count":0,"parent_topic_id":null,"parent_comment_id":null,"aggregated_at":"2026-09-23 23:40:18","created_at":"2026-09-23 23:40:18","updated_at":"2026-10-21 05:23:11","last_pipeline_at":"2026-09-24T01:40:18.222Z","cover_image_url":null,"visibility_policy":"public","permission_policy":"public_read","index_policy":"index","training_policy":"disallow","license_policy":"view_only","policy_version":1,"is_initiator":false,"can_edit_topic":false,"can_manage_media":false,"initiator_name":"AI 热点编辑部","published_at":"2026-09-23 23:40:18"},"topicTree":{"root":{"id":"71032999-4b4e-4b7a-a56d-107e869988df","title":"Mercury 2.5 LLM hits 770 tokens per second"},"sourceLayer":null,"isRoot":true},"sources":[{"id":"e27756fe-b25a-45af-aa25-adf7dbbe45cf","title":"Mercury 2.5 LLM hits 770 tokens per second","url":"https://artificialanalysis.ai/models/mercury-2-5","publisher":"Retro_Dev","published_at":"2026-09-23T22:16:19Z","type":"json","description":"","source_kind":"rss","canonical_source_url":""}],"modelAnswers":[{"id":"6ccb3beb-91d9-4583-af79-a6c2388a9e71","model_id":"3e37ec80-3eb9-4d78-9dd4-b5fa2206be5a","model_name":"Kimi-K3","provider":"alibaba","version":"kimi-k3","summary":"据S1标题，Mercury 2.5大语言模型推理速度达每秒770个token，消息由Retro_Dev发布并链接至第三方评测平台Artificial Analysis的模型页面。目前仅有标题级信息，测试条件、数据口径与官方细节均缺失，真实性尚待确认。","answer":"来源事实层面：当前唯一可用材料是S1的标题——『Mercury 2.5 LLM hits 770 tokens per second』，由Retro_Dev于2026-09-23发布，URL指向第三方模型评测平台Artificial Analysis的Mercury 2.5页面。因此可确认的事实仅是：有账号以标题形式宣称Mercury 2.5达到每秒770个token的速度；页面正文、测试条件、对比数据均未提供。 模型归纳与背景（非来源事实，属推断）：Mercury系列在业内以扩散式（diffusion）语言模型路线著称，主打远高于传统自回归模型的生成速度，因此770 tokens/s这一量级在技术上具备合理性，但这是基于既有知识的推测，不能视为本条来源的结论。链接指向Artificial Analysis这类第三方基准平台，暗示该数字可能来自标准化跑分而非厂商自报，可信度相对更高，但这同样属于推测。 意义解读：若数据属实，每秒770 token的输出速度对延迟敏感场景（实时代码补全、交互式Agent、语音助手、高并发API服务）有实际价值，也可能压低单位推理成本。但速度不等于质量，在缺乏准确率、上下文长度、硬件配置与价格信息的情况下，无法评估其综合竞争力，建议等待页面详情或官方公告后再下结论。","citations":[{"source_id":"e27756fe-b25a-45af-aa25-adf7dbbe45cf","title":"Mercury 2.5 LLM hits 770 tokens per second","url":"https://artificialanalysis.ai/models/mercury-2-5"}],"generated_at":"2026-09-23 23:57:19","entities":[{"name":"Mercury 2.5","type":"product","confidence":0.95},{"name":"Artificial Analysis","type":"organization","confidence":0.7},{"name":"Retro_Dev","type":"person","confidence":0.5}],"uncertainties":["无法访问S1页面正文，无法确认770 tokens/s的测试条件（硬件型号、批量大小、量化方式、输入输出长度等）","无法确认该数值是单用户输出速度还是聚合吞吐量，也未说明与模型质量评测的关系","发布时间2026-09-23超出本模型知识范围，无法核实Mercury 2.5的开发方、参数规模、许可与定价","Retro_Dev身份不明，可能为转载分享而非官方发布渠道","标题未给出对比基线，无法判断该速度在同类模型中的相对水平"]},{"id":"1bc69e92-957e-4f69-a035-a2be1864a293","model_id":"43bd6f33-1b6d-422d-8673-f10a76194d86","model_name":"智谱GLM5.2","provider":"bytedance","version":"glm-5-2-260617","summary":"根据Retro_Dev于2026年9月23日发布的信息，Mercury 2.5大语言模型运行速度达到每秒770个Token，显示出AI应用在推理效率上的最新突破。","answer":"【来源事实】根据 Retro_Dev 于2026年9月23日发布的信息，Mercury 2.5 大语言模型（LLM）的运行处理速度达到了每秒770个Token [S1]。 【模型归纳】在AI应用领域，大语言模型的推理速度直接关系到用户体验和应用场景的拓展。每秒770个Token的生成速度对于文本生成、实时对话和自动化内容创作等应用而言，是一个相当高的性能指标。这种速度的突破通常意味着底层架构或推理引擎进行了显著优化，能够大幅降低用户等待时间，为需要低延迟响应的复杂AI应用提供技术支撑。这也反映出当前AI行业发展不仅关注模型参数规模的增长，也高度注重推理效率和实际部署成本的优化。 【不确定性】仅基于当前来源，无法确认该速度测试的具体硬件环境、批处理大小、输入输出长度以及模型的具体参数量。此外，该数据由 Retro_Dev 发布，其测试的客观性和可复现性仍需更多第三方独立测试来验证。实际商业部署中能否保持这一峰值速度也具有不确定性。","citations":[{"source_id":"e27756fe-b25a-45af-aa25-adf7dbbe45cf","title":"Mercury 2.5 LLM hits 770 tokens per second","url":"https://artificialanalysis.ai/models/mercury-2-5"}],"generated_at":"2026-09-23 23:54:09","entities":[{"name":"Mercury 2.5","type":"product","confidence":0.95},{"name":"Mercury","type":"brand","confidence":0.85},{"name":"Retro_Dev","type":"organization","confidence":0.8},{"name":"LLM","type":"topic","confidence":0.9}],"uncertainties":["测试每秒770个Token的具体硬件环境和推理条件尚不明确","Retro_Dev作为发布方的具体背景及测试方法的可复现性无法确认","Mercury 2.5模型的具体参数规模和开发企业尚未披露"]},{"id":"4f1e0cb0-1722-40a9-9543-d108a88389d4","model_id":"42dbe57c-56d8-47c9-82ae-1988ea905f8e","model_name":"通义千问 (Qianwen)","provider":"alibaba","version":"qwen3.8-27b","summary":"来源S1标题称 Mercury 2.5 LLM 达770 tokens/s，由 Retro_Dev 发布并指向 Artificial Analysis 页面。可确认的是其关注生成速度；因缺少摘要和正文，无法判断测试硬件、吞吐类型、模型质量、价格及具体应用场景与性能边界。","answer":"根据唯一来源 S1，发布方 Retro_Dev 的时间为 2026-09-23T22:16:19Z，URL 指向 artificialanalysis.ai/models/mercury-2-5，标题为“Mercury 2.5 LLM hits 770 tokens per second”。来源中未提供热点摘要或正文，因此可确认的来源事实仅限于标题信息与元数据：存在一个名为 Mercury 2.5 的 LLM，且标题声称其速度达到每秒 770 个 tokens。模型可谨慎归纳为：该热点属于 AI 应用中的大语言模型推理效率议题，核心看点是 Mercury 2.5 的 token 生成速度。若该数字属实，770 tokens/s 表示较高的生成速率，可能对流式对话、实时辅助、批量内容生成等体验产生影响；但由于缺少测试说明，不能将其视为模型质量、准确性、成本或生产可用性的权威结论。不确定项包括：速度指标是峰值还是均值、对应什么硬件和上下文长度、是否考虑延迟和批处理、是否由官方或 Artificial Analysis 实测，以及 Mercury 2.5 的厂商、参数、上下文和价格。","citations":[{"source_id":"e27756fe-b25a-45af-aa25-adf7dbbe45cf","title":"Mercury 2.5 LLM hits 770 tokens per second","url":"https://artificialanalysis.ai/models/mercury-2-5"}],"generated_at":"2026-09-23 23:53:31","entities":[{"name":"Mercury 2.5","type":"product","confidence":0.95},{"name":"LLM","type":"topic","confidence":0.9},{"name":"Retro_Dev","type":"person","confidence":0.55},{"name":"Artificial Analysis","type":"organization","confidence":0.7},{"name":"AI应用","type":"topic","confidence":0.9}],"uncertainties":["来源只提供标题、发布方、时间和 URL，未提供正文或摘要，无法确认 770 tokens/s 的测试条件、硬件、上下文长度、批处理大小和延迟。","无法确认该速度是峰值吞吐、平均吞吐、官方宣称、第三方基准还是用户提交数据。","无法确认 Mercury 2.5 的开发商、参数规模、上下文长度、许可证、价格、准确性和实际应用场景。","无法确认 Retro_Dev 是个人账号、品牌账号还是机构账号。"]}],"modelInsight":{"common_points":[],"disagreements":[],"evidence_refs":[]},"entities":[{"id":"a735afa8-7595-4ebc-b0b4-f428e5d8d54b","name":"Mercury 2.5","type":"product","confidence":0.95,"visibility_score":100,"snapshot_count":3,"mentioned_count":3,"cited_count":3},{"id":"20f28f3b-6d1e-434b-bbce-f3fa96883eab","name":"LLM","type":"topic","confidence":0.9,"visibility_score":100,"snapshot_count":2,"mentioned_count":2,"cited_count":2},{"id":"29501d97-7888-4808-8811-7e4fd7439be0","name":"AI应用","type":"topic","confidence":0.9,"visibility_score":40,"snapshot_count":1,"mentioned_count":0,"cited_count":1},{"id":"1dcb5f2e-e3be-4985-8c10-c6ccb5fdc817","name":"Mercury","type":"brand","confidence":0.85,"visibility_score":100,"snapshot_count":1,"mentioned_count":1,"cited_count":1},{"id":"9fc69154-1d21-4803-abe4-6d2dd2a32468","name":"Artificial Analysis","type":"organization","confidence":0.7,"visibility_score":100,"snapshot_count":2,"mentioned_count":2,"cited_count":2},{"id":"e24386df-1aee-4689-bea3-7ee6a5bcde59","name":"Retro_Dev","type":"person","confidence":0.55,"visibility_score":100,"snapshot_count":3,"mentioned_count":3,"cited_count":3}],"modelStats":[{"answer_id":"6ccb3beb-91d9-4583-af79-a6c2388a9e71","model_id":"3e37ec80-3eb9-4d78-9dd4-b5fa2206be5a","model_name":"Kimi-K3","provider":"alibaba","generated_at":"2026-09-23 23:57:19","citation_count":1,"entity_count":3,"uncertainty_count":5},{"answer_id":"1bc69e92-957e-4f69-a035-a2be1864a293","model_id":"43bd6f33-1b6d-422d-8673-f10a76194d86","model_name":"智谱GLM5.2","provider":"bytedance","generated_at":"2026-09-23 23:54:09","citation_count":1,"entity_count":4,"uncertainty_count":3},{"answer_id":"4f1e0cb0-1722-40a9-9543-d108a88389d4","model_id":"42dbe57c-56d8-47c9-82ae-1988ea905f8e","model_name":"通义千问 (Qianwen)","provider":"alibaba","generated_at":"2026-09-23 23:53:31","citation_count":1,"entity_count":5,"uncertainty_count":4}],"topicAssets":[],"relatedTopics":[{"id":"716acce4-331a-428c-be53-86ef162d7837","title":"假牙粘附剂有股松油味","category":"AI与科技","heat_score":73,"discussion_volume":14,"summary":"假牙粘附剂有松油味或异味，主要是因为部分“天然”品牌使用松香树脂作为粘合基底，带来浓郁松木气味；同时也可能源于配方中的樟脑、薄荷等抑菌剂，或酒精溶剂的挥发。此外，若保存不当导致管口细菌滋生，也会产生严重的腐败异味。","updated_at":"2026-10-21 07:40:18"},{"id":"7acb0c81-6aee-4165-b80c-4968d199d0ab","title":"茶船古道茉莉六堡茶茶饮","category":"AI与科技","heat_score":71,"discussion_volume":8,"summary":"识别品牌为'茶船古道'，产品为500mL无糖茉莉六堡茶，由广西华林食品受托生产。该饮品主打0糖0脂0卡，融合传统六堡茶底与现代健康理念。其市场定位覆盖追求健康的年轻群体及肠胃调理人群，参照无糖茶饮市场，预估售价在5-8元区间。","updated_at":"2026-10-21 19:05:57"},{"id":"f375d21f-57c2-4780-982e-98068eb04bf0","title":"中国广电5G卡适合在地方合作吗","category":"AI与科技","heat_score":71,"discussion_volume":13,"summary":"广电5G卡在地方合作中具有双面性。对于政企数字化项目（如安防、专网），其政策支持力度大且成本较低，较为适合��但对于纯个人号卡的渠道代理，虽然初期具备价格优势和低门槛，但面临监管收紧和售后服务不完善的风险，需谨慎评估。","updated_at":"2026-10-21 07:28:35"},{"id":"e0220c65-8cb4-4da8-b025-d11f63782d3b","title":"哪吒汽车重组是不是黄了？","category":"AI与科技","heat_score":65,"discussion_volume":5,"summary":"根据联网检索结果，以下是我整理的汇总分析： ```json { \"summary\": \"哪吒汽车重组并未\"黄了\"，而是正处于破产重整的关键推进阶段。2026年9月11日，合众新能源第四次债权人会议披露《重整计划（草案）》，浙江太乙圣莲拟出资30亿元获约70.62%股权。债权人线上表决截止日为10月12日，其后仍需嘉兴中院裁定批准方可生效。此前2026年3月原方案曾因未按期提交被法院裁定转入清算，本次为重新推进的重整方案。\", \"answer\": \"哪吒汽车的重组目前并未彻底失","updated_at":"2026-10-21 07:27:49"},{"id":"0bb81ecd-03d0-4e58-8bd9-c72a0195e9ad","title":"现在网络游戏的环境真没法杜绝外挂的使用吗？","category":"AI与科技","heat_score":56,"discussion_volume":9,"summary":"彻底杜绝网络游戏外挂目前看来在技术和客观层面均不现实。主要原因包括：黑客黑产已形成高利润的工业化链条，不断推动技术迭代；以DMA（直接内存访问）为代表的硬件外挂突破了软件检测的物理边界，导致传统反作弊手段失效；且客户端代码完全受控于玩家端的局限性使得“绝对安全”无法实现。尽管有kernel级别防护和法律打击，但只能达到“抑制”而非“根除”。","updated_at":"2026-10-09 09:42:14"},{"id":"9e4ddbfd-6428-4fe3-82ae-94b711266fad","title":"Trump says US will henceforth call AI 'super intelligence'","category":"AI与科技","heat_score":41,"discussion_volume":7,"summary":"两家模型均援引外媒报道称，特朗普宣称美国将把“AI”更名为“超级智能（SI）”，并可能用于官方文件。但截至当前，均无白宫正式文件或行政命令确认，事件处于“媒体宣称、官方未证”阶段，真实性与具体执行细节仍有待核实。两则解读信息高度一致，未呈现实质差异。","updated_at":"2026-10-01 08:30:05"},{"id":"hot-demo-001","title":"AI Agent 正在进入企业工作流","category":"AI与科技","heat_score":40,"discussion_volume":6,"summary":"企业开始把 AI Agent 从单点聊天工具接入真实业务流程，热点讨论集中在可靠性、成本与自动化边界。","updated_at":"2026-10-21 10:47:28"},{"id":"f56154db-f68c-488f-bc54-0e9a50318cfa","title":"GPT-6 Sol (Max) Intelligence, Performance and Price Analysis","category":"AI与科技","heat_score":40,"discussion_volume":4,"summary":"综合各模型解读，可确认Artificial Analysis平台存在由theanonymousone于2026-09-22发布的《GPT-6 Sol (Max) 智能、性能与价格分析》报告，标题含(Max)。Kimi-K3还提到同日存在GPT-6 Luna及三份Claude Opus 5.5相关报告，但其他模型未提及。所有模型均表示无法获取具体基准得分、性能数据与定价信息，需查阅原始页面核实。各模型对核心事实的描述高度一致，未发现实质性分歧，仅信息详略不同。未确定项包括具体基准结果、性能指标、价格、对比结论及与其他模型的关系。","updated_at":"2026-09-30 05:33:22"}],"canonical_url":"http://www.aigc.guru/hot/topic/71032999-4b4e-4b7a-a56d-107e869988df","discussion":null}}