{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-10-24T05:00:40.349Z","headline":"Prime Intellect 发布 Prime Inference：面向开源前沿模型的 Serverless 与预留推理服务","description":"Prime Intellect 推出推理服务平台 Prime Inference，为开源前沿模型提供 serverless 端点和预留容量，上线前内部日处理近 1 万亿 token，主要来自 RL rollout、合成数据生成、评测和编码智能体。","url":"https://www.aioga.com/news/tyuegh8ybpndk7enxfhwlmkhx/","mainEntityOfPage":"https://www.aioga.com/news/tyuegh8ybpndk7enxfhwlmkhx/","datePublished":"2026-10-03T05:37:18.000Z","dateModified":"2026-10-03T05:37:18.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.marktechpost.com/2026/10/02/prime-intellect-launches-prime-inference-serverless-and-reserved-serving-for-frontier-open-models","https://aihot.news/items/tyuegh8ybpndk7enxfhwlmkhx"],"canonicalUrl":"https://www.aioga.com/news/tyuegh8ybpndk7enxfhwlmkhx/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：Prime Intellect 推出推理服务平台 Prime Inference，为开源前沿模型提供 serverless 端点和预留容量，上线前内部日处理近 1 万亿 token，主要来自 RL Aioga 将其归入「行业动态」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/tyuegh8ybpndk7enxfhwlmkhx/","dateCreated":"2026-10-03T05:37:18.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"MarkTechPost source article","url":"https://www.marktechpost.com/2026/10/02/prime-intellect-launches-prime-inference-serverless-and-reserved-serving-for-frontier-open-models","datePublished":"2026-10-03T05:37:18.000Z","provider":{"@type":"Organization","name":"MarkTechPost","url":"https://www.marktechpost.com/2026/10/02/prime-intellect-launches-prime-inference-serverless-and-reserved-serving-for-frontier-open-models"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.news/items/tyuegh8ybpndk7enxfhwlmkhx","datePublished":"2026-10-03T05:37:18.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.news/items/tyuegh8ybpndk7enxfhwlmkhx"}}],"aggregationSource":"MarkTechPost","originalPublisher":{"name":"MarkTechPost","url":"https://www.marktechpost.com/2026/10/02/prime-intellect-launches-prime-inference-serverless-and-reserved-serving-for-frontier-open-models"},"geoDeepAnswer":null,"article":{"id":"tyuegh8ybpndk7enxfhwlmkhx","slug":"tyuegh8ybpndk7enxfhwlmkhx","url":"https://www.aioga.com/news/tyuegh8ybpndk7enxfhwlmkhx/","title":"Prime Intellect 发布 Prime Inference：面向开源前沿模型的 Serverless 与预留推理服务","title_en":"","summary":"Prime Intellect 推出推理服务平台 Prime Inference，为开源前沿模型提供 serverless 端点和预留容量，上线前内部日处理近 1 万亿 token，主要来自 RL rollout、合成数据生成、评测和编码智能体。","source":"MarkTechPost","sourceUrl":"https://www.marktechpost.com/2026/10/02/prime-intellect-launches-prime-inference-serverless-and-reserved-serving-for-frontier-open-models","aiHotUrl":"https://aihot.news/items/tyuegh8ybpndk7enxfhwlmkhx","publishedAt":"2026-10-03T05:37:18.000Z","category":"行业动态","score":58,"selected":false,"articleBody":["Prime Intellect has launched Prime Inference：https://www.primeintellect.ai/blog/prime-inference, a serving platform for frontier open-source models. It offers serverless endpoints and reserved capacity on Prime’s own GPUs across multiple datacenters. Before public release, it processed nearly a trillion tokens per day internally. That traffic came from RL rollouts, synthetic data generation, evaluations and long-running coding agents.","Prime Inference is the serving layer of Prime Intellect’s open training stack. The company already ships post-training tools such as prime-rl, verifiers and sandboxes. Serving closes that loop: deployed models generate production traces that can feed back into training. Prime reports its GLM-5.3：https://huggingface.co/zai-org/GLM-5.3 endpoint ranks among the fastest on OpenRouter. It also cites a near-zero tool-call error rate and 100% uptime since launch.","The stack combines NVIDIA Dynamo：https://github.com/ai-dynamo/dynamo, vLLM：https://github.com/vllm-project/vllm, Mooncake：https://github.com/kvcache-ai/Mooncake and FlashInfer：https://github.com/flashinfer-ai/flashinfer. It was built with Inferact and NVIDIA, and fixes are contributed upstream.","The target workload is agentic. A typical agent turn adds about 6K tokens to a 140K-token prompt. Prime benchmarks this mix with SemiAnalysis AgentX：https://inferencex.semianalysis.com/agentx, and injected cold arrivals.","Prefill/decode disaggregation : Prefill and decode run on separate GPU groups. Dynamo handles routing, and vLLM runs the model on each group. Decoders pull computed KV through NIXL：https://github.com/ai-dynamo/nixl. Prime reports nearly 40% lower p90 inter-token latency in its tests.","Cache-aware routing :Dynamo’s KV-aware router weighs cached prefix overlap against queued work. Sessions stay on the same decoder between turns. Mooncake adds a second KV tier in host DRAM.","The interactivity target was 100 end-to-end tokens per second per user. At that bar, a 1:4 prefill/decode ratio served the most users. It reached 66 sessions per prefill group at 101 tok/s per user and 100 output tok/s per GPU.","Agents fail when tool calls carry wrong names or broken arguments. Prime Intellect’s team contributed a structural-tag builder to Dynamo for GLM’s tool format. vLLM then uses xgrammar：https://github.com/mlc-ai/xgrammar to mask tokens that violate the tool schema. The team also fixed parsing bugs, including being decoded into inside code.","Competitor prices verified October 2, 2026. Tracker figures come from ComputePrices, a third-party price tracker.","Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us ：https://forms.gle/MJjjVDPS7whH8Ngs6","Datalab Introduces OmniExtractBench to Fix Bias and Opacity in Extraction Benchmarks ：https://pxllnk.co/hxplrq","Practitioner-first AI/ML news and analysis, read by 1M+ developers and researchers every month."],"articleImages":[{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/10/blog1990-3-100x70.png","alt":"Meta, OpenAI and Uber Just Taught AI Agents to Talk First. What About When to Stay Quiet?","afterParagraph":9,"url":"/media/articles/tyuegh8ybpndk7enxfhwlmkhx/e274b5376c9b77bb.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/10/blog1990-2-100x70.png","alt":"IBM Brings Bob to Self-Hosted and Air-Gapped Environments: Agentic Software Development Without Moving Your Code","afterParagraph":9,"url":"/media/articles/tyuegh8ybpndk7enxfhwlmkhx/efe1daf16123c2e9.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/10/blog1990-1-100x70.png","alt":"Microsoft AI Releases MAI-Transcribe-2-Streaming: #1 Real-Time Speech-to-Text Model on Artificial Analysis","afterParagraph":9,"url":"/media/articles/tyuegh8ybpndk7enxfhwlmkhx/91356f05c5921a07.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/10/blog1990-100x70.png","alt":"Decision AI Models Explained: TypeSafe Jev vs Fastino GLiDE, GLiNER2.5-Decide and Open-Source Competitors","afterParagraph":9,"url":"/media/articles/tyuegh8ybpndk7enxfhwlmkhx/a0d19063de194d1d.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/10/blog149-4-100x70.png","alt":"NVIDIA Announces DGX Spark 64GB","afterParagraph":9,"url":"/media/articles/tyuegh8ybpndk7enxfhwlmkhx/c578f443c64764e7.webp"}],"mediaStatus":"ok","articleBodyZh":["Prime Intellect 已推出 Prime Inference：https://www.primeintellect.ai/blog/prime-inference，这是一个面向前沿开源模型的服务平台。它提供无服务器端点和 Prime 自有 GPU 的预留容量，覆盖多个数据中心。在公开发布前，它每天内部处理了近一万亿个 token。该流量来自 RL 回滚、合成数据生成、评估以及长时间运行的编码代理。","Prime Inference 是 Prime Intellect 开放训练堆栈的服务层。公司已经提供训练后工具，如 prime-rl、验证器和沙盒。服务层完成了这一闭环：部署的模型生成的生产轨迹可以反馈回训练环节。Prime 报告其 GLM-5.3：https://huggingface.co/zai-org/GLM-5.3 端点在 OpenRouter 上排名最快之一。它还提到几乎为零的工具调用错误率以及自发布以来 100% 的正常运行时间。","该堆栈结合了 NVIDIA Dynamo：https://github.com/ai-dynamo/dynamo、vLLM：https://github.com/vllm-project/vllm、Mooncake：https://github.com/kvcache-ai/Mooncake 和 FlashInfer：https://github.com/flashinfer-ai/flashinfer。它是与 Inferact 和 NVIDIA 一起构建的，修复内容也贡献给了上游项目。","目标工作负载是代理驱动型。一个典型的代理回合会向 140K token 的提示中增加约 6K token。Prime 通过 SemiAnalysis AgentX：https://inferencex.semianalysis.com/agentx，以及注入的冷启动请求来基准测试这种混合负载。","预填充/解码拆分：预填充和解码在不同的 GPU 组上运行。Dynamo 负责路由，vLLM 在每个组上运行模型。解码器通过 NIXL：https://github.com/ai-dynamo/nixl 拉取计算好的 KV。Prime 报告其测试中的 p90 token 间延迟降低了近 40%。","缓存感知路由：Dynamo 的 KV 感知路由器会权衡缓存前缀重叠与排队任务。会话在回合之间保持在同一解码器上。Mooncake 在主机 DRAM 中增加了第二级 KV。","互动性目标是每位用户每秒 100 个端到端 token。在该指标下，1:4 的预填充/解码比率服务了最多用户。它在每个预填充组达到 66 个会话，每位用户 101 tok/s，每个 GPU 输出 100 tok/s。","当工具调用使用错误的名称或损坏的参数时，代理会失败。Prime Intellect 的团队为 Dynamo 的 GLM 工具格式贡献了一个结构标签构建器。vLLM 然后使用 xgrammar：https://github.com/mlc-ai/xgrammar 来屏蔽违反工具模式的标记。团队还修复了解析错误，包括被解码到内部代码中。","竞争对手价格已于 2026 年 10 月 2 日验证。跟踪数据来自 ComputePrices，一家第三方价格跟踪工具。","需要与我们合作以推广您的 GitHub 仓库、Hugging Face 页面、产品发布或网络研讨会等吗？请联系：https://forms.gle/MJjjVDPS7whH8Ngs6","Datalab 推出了 OmniExtractBench，以解决提取基准中的偏差和不透明问题：https://pxllnk.co/hxplrq","面向从业者的 AI/ML 新闻和分析，每月有超过 100 万开发者和研究人员阅读。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：Prime Intellect 推出推理服务平台 Prime Inference，为开源前沿模型提供 serverless 端点和预留容量，上线前内部日处理近 1 万亿 token，主要来自 RL Aioga 将其归入「行业动态」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-10-24T05:04:22.866Z","sourceHash":"426e0779c176c724","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["行业动态","MarkTechPost"],"translations":{"zh-CN":{"title":"Prime Intellect 发布 Prime Inference：面向开源前沿模型的 Serverless 与预留推理服务","summary":"Prime Intellect 推出推理服务平台 Prime Inference，为开源前沿模型提供 serverless 端点和预留容量，上线前内部日处理近 1 万亿 token，主要来自 RL rollout、合成数据生成、评测和编码智能体。","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect 发布 Prime Inference：面向开源前沿模型的 Serverless 与预留推理服务 - Aioga AI资讯","description":"Prime Intellect 推出推理服务平台 Prime Inference，为开源前沿模型提供 serverless 端点和预留容量，上线前内部日处理近 1 万亿 token，主要来自 RL rollout、合成数据生成、评测和编码智能体。","url":"https://www.aioga.com/news/tyuegh8ybpndk7enxfhwlmkhx/","articleBody":["Prime Intellect 已推出 Prime Inference：https://www.primeintellect.ai/blog/prime-inference，这是一个面向前沿开源模型的服务平台。它提供无服务器端点和 Prime 自有 GPU 的预留容量，覆盖多个数据中心。在公开发布前，它每天内部处理了近一万亿个 token。该流量来自 RL 回滚、合成数据生成、评估以及长时间运行的编码代理。","Prime Inference 是 Prime Intellect 开放训练堆栈的服务层。公司已经提供训练后工具，如 prime-rl、验证器和沙盒。服务层完成了这一闭环：部署的模型生成的生产轨迹可以反馈回训练环节。Prime 报告其 GLM-5.3：https://huggingface.co/zai-org/GLM-5.3 端点在 OpenRouter 上排名最快之一。它还提到几乎为零的工具调用错误率以及自发布以来 100% 的正常运行时间。","该堆栈结合了 NVIDIA Dynamo：https://github.com/ai-dynamo/dynamo、vLLM：https://github.com/vllm-project/vllm、Mooncake：https://github.com/kvcache-ai/Mooncake 和 FlashInfer：https://github.com/flashinfer-ai/flashinfer。它是与 Inferact 和 NVIDIA 一起构建的，修复内容也贡献给了上游项目。","目标工作负载是代理驱动型。一个典型的代理回合会向 140K token 的提示中增加约 6K token。Prime 通过 SemiAnalysis AgentX：https://inferencex.semianalysis.com/agentx，以及注入的冷启动请求来基准测试这种混合负载。","预填充/解码拆分：预填充和解码在不同的 GPU 组上运行。Dynamo 负责路由，vLLM 在每个组上运行模型。解码器通过 NIXL：https://github.com/ai-dynamo/nixl 拉取计算好的 KV。Prime 报告其测试中的 p90 token 间延迟降低了近 40%。","缓存感知路由：Dynamo 的 KV 感知路由器会权衡缓存前缀重叠与排队任务。会话在回合之间保持在同一解码器上。Mooncake 在主机 DRAM 中增加了第二级 KV。","互动性目标是每位用户每秒 100 个端到端 token。在该指标下，1:4 的预填充/解码比率服务了最多用户。它在每个预填充组达到 66 个会话，每位用户 101 tok/s，每个 GPU 输出 100 tok/s。","当工具调用使用错误的名称或损坏的参数时，代理会失败。Prime Intellect 的团队为 Dynamo 的 GLM 工具格式贡献了一个结构标签构建器。vLLM 然后使用 xgrammar：https://github.com/mlc-ai/xgrammar 来屏蔽违反工具模式的标记。团队还修复了解析错误，包括被解码到内部代码中。","竞争对手价格已于 2026 年 10 月 2 日验证。跟踪数据来自 ComputePrices，一家第三方价格跟踪工具。","需要与我们合作以推广您的 GitHub 仓库、Hugging Face 页面、产品发布或网络研讨会等吗？请联系：https://forms.gle/MJjjVDPS7whH8Ngs6","Datalab 推出了 OmniExtractBench，以解决提取基准中的偏差和不透明问题：https://pxllnk.co/hxplrq","面向从业者的 AI/ML 新闻和分析，每月有超过 100 万开发者和研究人员阅读。"]},"en":{"title":"Prime Intellect Launches Prime Inference: Serverless and Reserved Inference Services for Open-Source Frontier Models","summary":"Prime Intellect has launched the reasoning service platform Prime Inference, providing serverless endpoints and reserved capacity for open-source cutting-edge models. Before going online, it internally processes nearly 1 trillion tokens per day, mainly from RL rollout, synthetic data generation, evaluation, and coding agents.","category":"Industry","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect Launches Prime Inference: Serverless and Reserved Inference Services for Open-Source Frontier Models - Aioga AI News","description":"Prime Intellect has launched the reasoning service platform Prime Inference, providing serverless endpoints and reserved capacity for open-source cutting-edge models. Before going...","url":"https://www.aioga.com/en/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T16:45:05.084Z"},"ja":{"title":"Prime Intellect、オープンソースの最先端モデル向けServerlessおよび予約制推論サービス「Prime Inference」を発表","summary":"Prime Intellect は推論サービスプラットフォーム Prime Inference を開始し、オープンソースの最先端モデルに対してサーバーレスのエンドポイントと予約容量を提供します。公開前には社内で日次でほぼ 1 兆トークンを処理しており、主に RL ロールアウト、合成データ生成、評価、コーディングエージェントに由来しています。","category":"業界動向","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect、オープンソースの最先端モデル向けServerlessおよび予約制推論サービス「Prime Inference」を発表 - Aioga AIニュース","description":"Prime Intellect は推論サービスプラットフォーム Prime Inference を開始し、オープンソースの最先端モデルに対してサーバーレスのエンドポイントと予約容量を提供します。公開前には社内で日次でほぼ 1 兆トークンを処理しており、主に RL ロールアウト、合成データ生成、評価、コーディングエージェントに由来しています。","url":"https://www.aioga.com/ja/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T16:45:26.590Z"},"ko":{"title":"Prime Intellect, Prime Inference 발표: 오픈 소스 최첨단 모델을 위한 서버리스 및 예약 추론 서비스","summary":"Prime Intellect가 추론 서비스 플랫폼 Prime Inference를 출시해 오픈소스 최첨단 모델에 서버리스 엔드포인트와 예약 용량을 제공한다. 출시 전 내부적으로 하루 처리량은 거의 1조 토큰에 달했으며, 주로 RL rollout, 합성 데이터 생성, 평가 및 코딩 에이전트에서 발생했다.","category":"업계 동향","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect, Prime Inference 발표: 오픈 소스 최첨단 모델을 위한 서버리스 및 예약 추론 서비스 - Aioga AI 뉴스","description":"Prime Intellect가 추론 서비스 플랫폼 Prime Inference를 출시해 오픈소스 최첨단 모델에 서버리스 엔드포인트와 예약 용량을 제공한다. 출시 전 내부적으로 하루 처리량은 거의 1조 토큰에 달했으며, 주로 RL rollout, 합성 데이터 생성, 평가 및 코딩 에이전트에서 발생했다.","url":"https://www.aioga.com/ko/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T16:45:26.903Z"},"es":{"title":"Prime Intellect lanza Prime Inference: servicios de inferencia sin servidor y reservados para modelos de vanguardia de código abierto","summary":"Prime Intellect lanza la plataforma de servicios de inferencia Prime Inference, que ofrece endpoints serverless y capacidad reservada para modelos de vanguardia de código abierto. Antes de su lanzamiento, procesaba internamente cerca de 1 billón de tokens al día, principalmente para rollouts de RL, generación de datos sintéticos, evaluación y agentes de codificación.","category":"Industria","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect lanza Prime Inference: servicios de inferencia sin servidor y reservados para modelos de vanguardia de código abierto - Aioga Noticias de IA","description":"Prime Intellect lanza la plataforma de servicios de inferencia Prime Inference, que ofrece endpoints serverless y capacidad reservada para modelos de vanguardia de código abierto....","url":"https://www.aioga.com/es/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T16:50:41.144Z"},"fr":{"title":"Prime Intellect a publié Prime Inference : services d'inférence serverless et réservés pour les modèles open source de pointe","summary":"Prime Intellect a lancé la plateforme de services de raisonnement Prime Inference, offrant des points de terminaison serverless et une capacité réservée pour les modèles de pointe open source. Avant son lancement, elle traitait en interne près de 1 000 milliards de tokens par jour, principalement issus du RL rollout, de la génération de données synthétiques, de l'évaluation et des agents de codage intelligents.","category":"Industrie","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect a publié Prime Inference : services d'inférence serverless et réservés pour les modèles open source de pointe - Aioga Actualités IA","description":"Prime Intellect a lancé la plateforme de services de raisonnement Prime Inference, offrant des points de terminaison serverless et une capacité réservée pour les modèles de pointe...","url":"https://www.aioga.com/fr/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T16:50:48.284Z"},"de":{"title":"Prime Intellect veröffentlicht Prime Inference: Serverless- und reservierte Inferenzdienste für hochmoderne Open-Source-Modelle.","summary":"Prime Intellect hat die Inferenz-Servic-Plattform Prime Inference eingeführt, die serverlose Endpunkte und reservierte Kapazitäten für Open-Source-Frontmodell bietet. Vor dem Launch wurden intern täglich fast 1 Billion Token verarbeitet, hauptsächlich aus RL-Rollouts, synthetischer Datengenerierung, Evaluierung und Codierungsagenten.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect veröffentlicht Prime Inference: Serverless- und reservierte Inferenzdienste für hochmoderne Open-Source-Modelle. - Aioga KI-News","description":"Prime Intellect hat die Inferenz-Servic-Plattform Prime Inference eingeführt, die serverlose Endpunkte und reservierte Kapazitäten für Open-Source-Frontmodell bietet. Vor dem Launc...","url":"https://www.aioga.com/de/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T16:51:16.964Z"},"pt-BR":{"title":"Prime Intellect lança Prime Inference: serviço de inferência sem servidor e reservado para modelos de ponta de código aberto","summary":"Prime Intellect lançou a plataforma de serviços de inferência Prime Inference, fornecendo endpoints serverless e capacidade reservada para modelos de ponta open source. Antes do lançamento, processou internamente quase 1 trilhão de tokens por dia, principalmente provenientes de RL rollout, geração de dados sintéticos, avaliação e agentes de codificação.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect lança Prime Inference: serviço de inferência sem servidor e reservado para modelos de ponta de código aberto - Aioga Notícias de IA","description":"Prime Intellect lançou a plataforma de serviços de inferência Prime Inference, fornecendo endpoints serverless e capacidade reservada para modelos de ponta open source. Antes do la...","url":"https://www.aioga.com/pt-BR/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T16:56:26.502Z"},"ru":{"title":"Prime Intellect представляет Prime Inference: бессерверные и резервируемые сервисы инференса для передовых моделей с открытым исходным кодом","summary":"Prime Intellect запускает платформу вывода Prime Inference, предоставляющую serverless-конечные точки и зарезервированные мощности для передовых моделей с открытым исходным кодом. До запуска внутренняя обработка достигала почти 1 триллиона токенов в день, в основном из RL rollout, генерации синтетических данных, оценки и кодирования агентов.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect представляет Prime Inference: бессерверные и резервируемые сервисы инференса для передовых моделей с открытым исходным кодом - Aioga Новости ИИ","description":"Prime Intellect запускает платформу вывода Prime Inference, предоставляющую serverless-конечные точки и зарезервированные мощности для передовых моделей с открытым исходным кодом....","url":"https://www.aioga.com/ru/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T16:55:59.254Z"},"ar":{"title":"أصدرت Prime Intellect خدمة Prime Inference: خدمة استدلال بدون خادم ومحجوزة لنماذج المصدر المفتوح المتقدمة","summary":"أطلقت شركة Prime Intellect منصة خدمات الاستدلال Prime Inference، لتوفير نقاط نهاية بدون خادم وسعة محجوزة للنماذج الرائدة مفتوحة المصدر، وقد処لت داخليًا ما يقرب من 1 تريليون توكن يوميًا قبل الإطلاق، التي تأتي أساسًا من RL rollout، وتوليد البيانات التركيبية، والتقييم، ووكلاء الترميز الذكي.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"أصدرت Prime Intellect خدمة Prime Inference: خدمة استدلال بدون خادم ومحجوزة لنماذج المصدر المفتوح المتقدمة - Aioga أخبار الذكاء الاصطناعي","description":"أطلقت شركة Prime Intellect منصة خدمات الاستدلال Prime Inference، لتوفير نقاط نهاية بدون خادم وسعة محجوزة للنماذج الرائدة مفتوحة المصدر، وقد処لت داخليًا ما يقرب من 1 تريليون توكن يوم...","url":"https://www.aioga.com/ar/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T16:56:34.258Z"},"hi":{"title":"Prime Intellect ने Prime Inference जारी किया: ओपन-सोर्स अत्याधुनिक मॉडलों के लिए सर्वरलेस और आरक्षित इन्फरेंस सेवाएँ।","summary":"Prime Intellect ने तर्क सेवा प्लेटफ़ॉर्म Prime Inference लॉन्च किया, जो ओपन-सोर्स अग्रिम मॉडल के लिए serverless एंडपॉइंट और आरक्षित क्षमता प्रदान करता है,上线前 आंतरिक रूप से करीब 1 ट्रिलियन टोकन प्रतिदिन संसाधित किए जाते थे, मुख्य रूप से RL रोलआउट, संश्लेषित डेटा निर्माण, मूल्यांकन और कोडिंग एजेंट से आते हैं।","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect ने Prime Inference जारी किया: ओपन-सोर्स अत्याधुनिक मॉडलों के लिए सर्वरलेस और आरक्षित इन्फरेंस सेवाएँ। - Aioga AI समाचार","description":"Prime Intellect ने तर्क सेवा प्लेटफ़ॉर्म Prime Inference लॉन्च किया, जो ओपन-सोर्स अग्रिम मॉडल के लिए serverless एंडपॉइंट और आरक्षित क्षमता प्रदान करता है,上线前 आंतरिक रूप से करीब 1 ट...","url":"https://www.aioga.com/hi/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T17:01:51.047Z"},"it":{"title":"Prime Intellect lancia Prime Inference: servizi di inferenza serverless e riservati per modelli open source all'avanguardia.","summary":"Prime Intellect ha lanciato la piattaforma di servizi di inferenza Prime Inference, che fornisce endpoint serverless e capacità riservata per modelli open source all'avanguardia. Prima del lancio, l'elaborazione interna giornaliera ha raggiunto quasi 1 trilione di token, provenienti principalmente da RL rollout, generazione di dati sintetici, valutazione e agenti di codifica intelligenti.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect lancia Prime Inference: servizi di inferenza serverless e riservati per modelli open source all'avanguardia. - Aioga Notizie IA","description":"Prime Intellect ha lanciato la piattaforma di servizi di inferenza Prime Inference, che fornisce endpoint serverless e capacità riservata per modelli open source all'avanguardia. P...","url":"https://www.aioga.com/it/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T17:01:50.290Z"},"nl":{"title":"Prime Intellect lanceert Prime Inference: serverloze en gereserveerde inferentiediensten voor open source cutting-edge modellen","summary":"Prime Intellect lanceert het inferentieplatform Prime Inference, dat serverless-eindpunten en gereserveerde capaciteit biedt voor toonaangevende open-sourcemodellen. Voor de lancering verwerkte het intern dagelijks bijna 1 biljoen tokens, voornamelijk afkomstig van RL-rollouts, het genereren van synthetische data, evaluaties en codeeragenten.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect lanceert Prime Inference: serverloze en gereserveerde inferentiediensten voor open source cutting-edge modellen - Aioga AI-nieuws","description":"Prime Intellect lanceert het inferentieplatform Prime Inference, dat serverless-eindpunten en gereserveerde capaciteit biedt voor toonaangevende open-sourcemodellen. Voor de lancer...","url":"https://www.aioga.com/nl/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T17:01:56.820Z"},"tr":{"title":"Prime Intellect, Prime Inference'i Yayınladı: Açık Kaynak Öncü Modeller İçin Sunucusuz ve Rezervli Çıkarım Hizmetleri","summary":"Prime Intellect, Prime Inference adında bir akıl yürütme hizmeti platformu başlattı ve açık kaynaklı ileri düzey modeller için serverless uç noktaları ve ayrılmış kapasite sağlıyor; yayına girmeden önce dahili olarak günde yaklaşık 1 trilyon token işliyordu, bunların çoğu RL rollout, sentetik veri oluşturma, değerlendirme ve kodlama ajanlarından geliyordu.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect, Prime Inference'i Yayınladı: Açık Kaynak Öncü Modeller İçin Sunucusuz ve Rezervli Çıkarım Hizmetleri - Aioga AI Haberleri","description":"Prime Intellect, Prime Inference adında bir akıl yürütme hizmeti platformu başlattı ve açık kaynaklı ileri düzey modeller için serverless uç noktaları ve ayrılmış kapasite sağlıyor...","url":"https://www.aioga.com/tr/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T17:07:20.491Z"},"vi":{"title":"Prime Intellect phát hành Prime Inference: Dịch vụ suy luận Serverless và dự trữ dành cho các mô hình tiên tiến mã nguồn mở","summary":"Prime Intellect ra mắt nền tảng dịch vụ suy luận Prime Inference, cung cấp các điểm cuối serverless và dung lượng dự trữ cho các mô hình tiên tiến nguồn mở, xử lý gần 1 nghìn tỷ token mỗi ngày nội bộ trước khi ra mắt, chủ yếu từ RL rollout, tạo dữ liệu tổng hợp, đánh giá và đại lý mã hóa.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect phát hành Prime Inference: Dịch vụ suy luận Serverless và dự trữ dành cho các mô hình tiên tiến mã nguồn mở - Tin tức AI Aioga","description":"Prime Intellect ra mắt nền tảng dịch vụ suy luận Prime Inference, cung cấp các điểm cuối serverless và dung lượng dự trữ cho các mô hình tiên tiến nguồn mở, xử lý gần 1 nghìn tỷ to...","url":"https://www.aioga.com/vi/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T17:07:52.917Z"},"id":{"title":"Prime Intellect merilis Prime Inference: Layanan inferensi tanpa server dan cadangan untuk model open-source terdepan","summary":"Prime Intellect meluncurkan platform layanan inferensi Prime Inference, yang menyediakan endpoint serverless dan kapasitas terreservasi untuk model-model mutakhir sumber terbuka. Sebelum peluncuran, platform ini memproses hampir 1 triliun token per hari untuk penggunaan internal, terutama untuk rollout RL, pembuatan data sintetis, evaluasi, dan agen pengodean.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect merilis Prime Inference: Layanan inferensi tanpa server dan cadangan untuk model open-source terdepan - Berita AI Aioga","description":"Prime Intellect meluncurkan platform layanan inferensi Prime Inference, yang menyediakan endpoint serverless dan kapasitas terreservasi untuk model-model mutakhir sumber terbuka. S...","url":"https://www.aioga.com/id/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T17:07:19.194Z"},"th":{"title":"Prime Intellect เปิดตัว Prime Inference: บริการการอนุมานแบบ Serverless และสำรองสำหรับโมเดลแนวหน้าแบบโอเพนซอร์ส","summary":"Prime Intellect เปิดตัวแพลตฟอร์มบริการการอนุมาน Prime Inference สำหรับโมเดลแนวหน้าที่เป็นโอเพนซอร์ส ให้บริการ serverless endpoints และความจุสำรอง ก่อนเปิดตัวมีการประมวลผลภายในวันละเกือบ 1 ล้านล้าน token ส่วนใหญ่มาจาก RL rollout การสร้างข้อมูลสังเคราะห์ การประเมินผล และตัวแทนรหัส","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect เปิดตัว Prime Inference: บริการการอนุมานแบบ Serverless และสำรองสำหรับโมเดลแนวหน้าแบบโอเพนซอร์ส - ข่าว AI Aioga","description":"Prime Intellect เปิดตัวแพลตฟอร์มบริการการอนุมาน Prime Inference สำหรับโมเดลแนวหน้าที่เป็นโอเพนซอร์ส ให้บริการ serverless endpoints และความจุสำรอง ก่อนเปิดตัวมีการประมวลผลภายในวันละ...","url":"https://www.aioga.com/th/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T17:13:04.307Z"},"pl":{"title":"Prime Intellect wypuszcza Prime Inference: bezserwerowe i zarezerwowane usługi inferencji dla otwartoźródłowych modeli najnowszej generacji","summary":"Prime Intellect uruchomił platformę usług rozumowania Prime Inference, oferującą bezserwerowe punkty końcowe i rezerwową pojemność dla otwartych modeli najnowszej generacji; przed uruchomieniem wewnętrznie przetwarzano prawie bilion tokenów dziennie, głównie pochodzących z RL rollout, generowania danych syntetycznych, oceniania i agentów kodujących.","category":"行业动态","source":"MarkTechPost","aggregationSource":"MarkTechPost","pageTitle":"Prime Intellect wypuszcza Prime Inference: bezserwerowe i zarezerwowane usługi inferencji dla otwartoźródłowych modeli najnowszej generacji - Aioga Wiadomości AI","description":"Prime Intellect uruchomił platformę usług rozumowania Prime Inference, oferującą bezserwerowe punkty końcowe i rezerwową pojemność dla otwartych modeli najnowszej generacji; przed...","url":"https://www.aioga.com/pl/news/tyuegh8ybpndk7enxfhwlmkhx/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"43ceb57febf0e0b8","translatedAt":"2026-10-03T17:13:10.198Z"}},"evidenceTier":"verified-news","reviewStatus":"automated-ingest","indexable":true,"editorialCover":""}}