{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T08:01:28.298Z","headline":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","description":"Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","url":"https://www.aioga.com/news/cmrmra2cn01t4bi7i4n1sq8vs/","mainEntityOfPage":"https://www.aioga.com/news/cmrmra2cn01t4bi7i4n1sq8vs/","datePublished":"2026-07-15T23:48:58.000Z","dateModified":"2026-07-15T23:48:58.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.marktechpost.com/2026/07/15/thinking-machines-lab-releases-inkling-a-975b-parameter-open-weights-multimodal-moe-with-41b-active-parameters-and-controllable-thinking-effort","https://aihot.virxact.com/items/cmrmra2cn01t4bi7i4n1sq8vs"],"canonicalUrl":"https://www.aioga.com/news/cmrmra2cn01t4bi7i4n1sq8vs/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。 Aioga 将其归入「模型更新」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/cmrmra2cn01t4bi7i4n1sq8vs/","dateCreated":"2026-07-15T23:48:58.000Z","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.com source article","url":"https://www.marktechpost.com/2026/07/15/thinking-machines-lab-releases-inkling-a-975b-parameter-open-weights-multimodal-moe-with-41b-active-parameters-and-controllable-thinking-effort","datePublished":"2026-07-15T23:48:58.000Z","provider":{"@type":"Organization","name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/15/thinking-machines-lab-releases-inkling-a-975b-parameter-open-weights-multimodal-moe-with-41b-active-parameters-and-controllable-thinking-effort"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrmra2cn01t4bi7i4n1sq8vs","datePublished":"2026-07-15T23:48:58.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrmra2cn01t4bi7i4n1sq8vs"}}],"aggregationSource":"MarkTechPost（RSS）","originalPublisher":{"name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/15/thinking-machines-lab-releases-inkling-a-975b-parameter-open-weights-multimodal-moe-with-41b-active-parameters-and-controllable-thinking-effort"},"article":{"id":"cmrmra2cn01t4bi7i4n1sq8vs","slug":"cmrmra2cn01t4bi7i4n1sq8vs","url":"https://www.aioga.com/news/cmrmra2cn01t4bi7i4n1sq8vs/","title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","title_en":"Thinking Machines Lab Releases Inkling： A 975B-Parameter Open-Weights Multimodal MoE With 41B Active Parameters And Controllable Thinking Effort","summary":"Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","source":"MarkTechPost（RSS）","sourceUrl":"https://www.marktechpost.com/2026/07/15/thinking-machines-lab-releases-inkling-a-975b-parameter-open-weights-multimodal-moe-with-41b-active-parameters-and-controllable-thinking-effort","aiHotUrl":"https://aihot.virxact.com/items/cmrmra2cn01t4bi7i4n1sq8vs","publishedAt":"2026-07-15T23:48:58.000Z","category":"模型更新","score":49,"selected":false,"articleBody":["Thinking Machines Lab just released Inkling：https://thinkingmachines.ai/model-card/inkling/ , their first model trained from scratch, weights are open, fine-tunable on Tinker. The lab pitches it as a base for customization.","Inkling is a Mixture-of-Experts transformer with 975B total parameters and 41B active. It supports a context window of up to 1M tokens. Pretraining covered 45 trillion tokens of text, images, audio, and video. Inputs accept text, images, and audio; output is UTF-8 text only.","The research team also previewed Inkling-Small, a 276B-parameter MoE with 12B active parameters. It matches or exceeds its larger sibling on many benchmarks, and its weights arrive once testing finishes. Because customization/finetuning is the key differentiator, the architecture matters here very much.","The model architecture includes a 66-layer decoder-only transformer with a sparse MoE feed-forward backbone. Each MoE layer holds 256 routed experts plus 2 shared experts. Six routed experts activate per token, and both shared experts activate on every token. A sigmoid-based router handles selection, using an auxiliary-loss-free load-balancing bias. Routed and shared scores are normalized jointly, then used to weight combined outputs. The MoE design largely follows DeepSeek-V3.","Attention departs from convention. Sliding-window and global layers interleave at a 5:1 ratio with 8 KV heads. Position uses a relative positional embedding rather than RoPE, which the lab reports extrapolates better. Short convolutions are applied after key and value projections, and on residual branch outputs.","Multimodality is encoder-free. Audio enters as dMel spectrograms, and images become 40×40 pixel patches through a four-layer hMLP. A lightweight embedding layer projects both, then the decoder processes them jointly with text tokens.","Training used Muon for large matrix weights and Adam for other parameters, on NVIDIA GB300 NVL72 systems. Post-training bootstrapped from SFT on synthetic data, including data generated by Kimi K2.5. Most compute went to asynchronous RL, scaled past 30M rollouts, improving log-linearly throughout. That RL run also produced the model’s main control surface.","During RL, the research team set effort by changing the system message and adjusting per-token cost. The model consequently learned to spend different token budgets on different rollouts. The release post sweeps effort from 0.2 to 0.99, and harnesses can set it directly. In transformers , the same control is exposed as a reasoning_effort argument with named levels.","The efficiency data is quite specific. Inkling spends one third as many tokens as Nemotron 3 Ultra for equal Terminal Bench 2.1 performance. Cost and latency become tunable per call, not fixed per model.","Alongside effort, the research team targeted trustworthiness directly.","All Inkling evals run at effort=0.99 and temperature 1.0, with a 256K trajectory limit for coding. Several scores are externally reported by Artificial Analysis. Against open-weights peers, the picture is quite competitive.","Inkling leads this open-weights group on FORTRESS Adversarial at 78.0%. It trails GLM 5.2 on Terminal Bench 2.1 by 18.9 points. It reports 73.5% on MMMU Pro and 91.4% on VoiceBench. It posts 1257 on Design Arena’s Agentic Web Dev leaderboard, a blinded human evaluation.","With numbers established, deployment becomes the practical question.","Two checkpoints ship. BF16 needs at least 2 TB aggregated VRAM: 8x NVIDIA B300 or 16x H200. NVFP4 drops that to at least 600 GB, running W4A4 on 4x B300 or W4A16 on 8x H200. Runtimes include SGLang, vLLM, TokenSpeed, Unsloth, and Hugging Face transformers .","OpenAI-compatible serving takes one command:","For fine-tuning, Inkling is live on Tinker with 64K and 256K context options. The research team also released tml-renderers for post-training with tool calls and multimodal inputs. Hosted APIs exist via TogetherAI, Fireworks, Modal, Databricks, and Baseten.","Given those constraints, three deployment patterns follow.","Taken together, the trade-offs are clear.","Asif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences."],"articleImages":[{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2019/06/Screen-Shot-2021-09-14-at-9.02.24-AM-300x300.png","alt":"","afterParagraph":17,"url":"/media/articles/cmrmra2cn01t4bi7i4n1sq8vs/787a6d54564e8e19.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/07/high-level-description-a-developer-focus_feCO3rqGV4ig6q7LaBcN2w_K5t5TwjZTPWy666HxX0epA-100x70.png","alt":"Patter SDK Guide to Building a Restaurant Booking Phone Agent with Dynamic Variables, Guardrails, Latency Dashboards, and Eval Checks","afterParagraph":18,"url":"/media/articles/cmrmra2cn01t4bi7i4n1sq8vs/c998049732b3b311.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/07/high-level-description-a-cartoonish-meme_mx9sd4-zUJ2CQbUyJxX95w_w7GS0nEmTc6N2_1EO2Snmg-100x70.png","alt":"Soofi Consortium Releases Soofi S 30B-A3B","afterParagraph":18,"url":"/media/articles/cmrmra2cn01t4bi7i4n1sq8vs/02ab67c4ea79309c.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/07/high-level-description-a-technical-blog-_jc7iI1-vWAa65GFgmnBhwQ_U7n28gZXRDeiQEimQU9cAw_cover-100x70.png","alt":"Building a Gin Config Controlled PyTorch Pipeline with Configurable MLP Variants, Cosine Scheduling, and Runtime Parameter Overrides","afterParagraph":18,"url":"/media/articles/cmrmra2cn01t4bi7i4n1sq8vs/7ac38241502a486f.webp"}],"mediaStatus":"ok","articleBodyZh":[],"translationStatus":"","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。 Aioga 将其归入「模型更新」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":"2026-07-23T08:10:16.893Z","sourceHash":"81dc14e5128de968","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["模型更新","MarkTechPost（RSS）"],"translations":{"zh-CN":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga AI资讯","description":"Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","url":"https://www.aioga.com/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"en":{"title":"Thinking Machines Lab Releases Inkling： A 975B-Parameter Open-Weights Multimodal MoE With 41B Active Parameters And Controllable Thinking Effort","summary":"Aioga tracks this update from MarkTechPost（RSS） under Models. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"Models","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab Releases Inkling： A 975B-Parameter Open-Weights Multimodal MoE With 41B Active Parameters And Controllable Thinking Effort - Aioga AI News","description":"Aioga tracks this update from MarkTechPost（RSS） under Models. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1","url":"https://www.aioga.com/en/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"ja":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aiogaは「モデル更新」の動きとして、MarkTechPost（RSS） からの更新を追跡しています。Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"モデル更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga AIニュース","description":"Aiogaは「モデル更新」の動きとして、MarkTechPost（RSS） からの更新を追跡しています。Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下","url":"https://www.aioga.com/ja/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"ko":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga는 MarkTechPost（RSS）의 업데이트를 모델 업데이트 흐름으로 추적합니다. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"모델 업데이트","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga AI 뉴스","description":"Aioga는 MarkTechPost（RSS）의 업데이트를 모델 업데이트 흐름으로 추적합니다. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下","url":"https://www.aioga.com/ko/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"es":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga sigue esta actualización de MarkTechPost（RSS） dentro de Modelos. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"Modelos","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga Noticias de IA","description":"Aioga sigue esta actualización de MarkTechPost（RSS） dentro de Modelos. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B ","url":"https://www.aioga.com/es/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"fr":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga suit cette mise à jour de MarkTechPost（RSS） dans la catégorie Modèles. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"Modèles","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga Actualités IA","description":"Aioga suit cette mise à jour de MarkTechPost（RSS） dans la catégorie Modèles. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transforme","url":"https://www.aioga.com/fr/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"de":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga KI-News","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/de/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"pt-BR":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga Notícias de IA","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/pt-BR/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"ru":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga Новости ИИ","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/ru/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"ar":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga أخبار الذكاء الاصطناعي","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/ar/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"hi":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga AI समाचार","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/hi/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"it":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga Notizie IA","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/it/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"nl":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga AI-nieuws","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/nl/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"tr":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga AI Haberleri","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/tr/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"vi":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Tin tức AI Aioga","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/vi/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"id":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Berita AI Aioga","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/id/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"th":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - ข่าว AI Aioga","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/th/news/cmrmra2cn01t4bi7i4n1sq8vs/"},"pl":{"title":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型","summary":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M token 上下文窗口及文本、图像、音频输入，核心差异化在于可控思考深度。","category":"模型更新","source":"MarkTechPost（RSS）","pageTitle":"Thinking Machines Lab 发布 Inkling：975B 参数开源多模态 MoE 模型 - Aioga Wiadomości AI","description":"Aioga tracks this update from MarkTechPost（RSS） under 模型更新. Thinking Machines Lab 于 7 月 15 日发布其首个从头训练的模型 Inkling，采用 Apache 2.0 协议开源完整权重。该模型为 975B 参数 MoE Transformer，41B 活跃参数，支持 1M ","url":"https://www.aioga.com/pl/news/cmrmra2cn01t4bi7i4n1sq8vs/"}}}}