
美國新創 Thinking Machines Lab 於11月12日開放 Inkling 2 Small 模型權重下載,採 Apache 2.0 授權,支援最高100萬 token 上下文與多模態輸入,NVFP4 版本至少需500GB GPU顯存,台灣企業與機關使用仍須評估資料安全與風險。
AI-generated summary
Thinking Machines Lab 是美國人工智慧新創,Inkling 2 Small 模型最早於11月5日發表,初期僅透過官方 Tinker 平台使用,待外部安全測試完成後於11月12日開放權重下載。
美國人工智慧新創 Thinking Machines Lab 11月12日正式開放 Inkling 2 Small 模型權重下載,採用 Apache 2.0 授權,允許開發者自行部署與微調。這款模型最早於11月5日發表,初期僅能透過官方 Tinker 平台使用,待外部安全測試完成後才開放權重。Inkling 2 Small 擁有5940億總參數(594B),每次推論約啟用170億參數(17B),支援最高100萬 token 上下文,主要支援 AI 代理、程式開發、資料分析及企業自動化應用。
Inkling 2 Small 採用混合專家(MoE)架構,具有48層 Transformer,每次處理文字時會從384個專家中選出8個參與運算,藉此減少推論所需的計算量。模型可同時理解文字與圖片,並以文字輸出結果,也支援工具呼叫及不同程度的推理設定。開發者能依需求將它用於聊天機器人、編碼助理或檢索增強生成(RAG)系統,不過新版本並未提供音訊輸入功能。
這款模型另一項特色是採用 NVIDIA NVFP4 低精度格式進行訓練,從預訓練到後訓練均以相同精度完成。Thinking Machines Lab 同時提供 NVFP4 與 MXFP8 兩種權重版本,其中 NVFP4 是官方建議的部署選擇,因為 MXFP8 在多項測試中僅提高約1至2個百分點,卻需要更多記憶體。根據模型卡,NVFP4 版本至少需要500GB GPU顯存,可使用兩張 NVIDIA B300 或四張 H200 執行;MXFP8 則需要至少800GB。儘管每次只啟用170億參數,完整模型權重仍須載入,因此並非一般消費級顯示卡就能輕易運行。
效能方面,官方公布的 BrowseComp 網頁搜尋測試取得90.2%,Finance Agent v2.1 金融代理測試則有67.5%,在同場比較的模型中均取得最高成績。SWEBench Pro v2 的 Full 測試達97.7%,但難度較高的 Hard 測試為86.3%,顯示不同測試條件下的表現仍有差異。另一項 AA-Omniscience v1.1 評測中,模型的非幻覺率為73.3%,代表在未能正確回答的情況下,較能避免憑空捏造內容,不能直接解讀成所有回答都有73.3%的正確率。
Inkling 2 Small 並非每項測試都領先。在 Artificial Analysis Intelligence Index v5.0 中,模型取得30.6分,略低於部分中國開放權重模型;Terminal-Bench 4.0 的成績則只有13.1%,與其他擅長終端機操作的模型仍有明顯差距。Thinking Machines Lab 也坦言,新模型在複雜工具選擇、內容組織與一般聊天體驗上還有改善空間,部分情境可能出現過度拒絕回答的問題。
值得注意的是,Inkling 2 Small 由美國公司開發,並採相對寬鬆的 Apache 2.0 授權,提供台灣企業與研究機構另一種非中國來源的模型選擇。不過,能免費下載或自行部署,不代表可直接用於所有政府業務。依照台灣行政院的生成式 AI 使用指引,公務機關仍須評估資料安全、個資保護與使用風險,機密文書及涉及未公開公務資訊的處理也有額外限制。若採封閉式地端部署,仍須先確認系統環境安全性,並遵守各機關的內部管理規範。
AI outlook — possibilities, not facts
更多開發者將下載並部署 Inkling 2 Small 模型用於 AI 代理、程式開發與企業自動化應用
Likely · Within weeks
台灣企業與研究機構將評估該模型在本地部署與應用的可行性
Possible · Within months

On November 10, the Federal Aviation Administration announced the expansion of the deployment of the AI air traffic management system SMART to 22 control centers across the United States. Three days later, a nationwide flight suspension incident occurred. Officials have not yet confirmed a direct relationship between the two. The accident resulted in the cancellation of more than 1,000 flights and the delay of nearly 10,000 flights. The Secretary of Homeland Security stated that no signs of malicious cyber attacks were found.

The US Federal Bureau of Investigation (FBI) issued a security alert on November 13, stating that the open source AI penetration testing tool ARTEX has been abused by hackers since August this year, attacking multiple industries such as finance, government, national defense, semiconductors, energy and technology. Some intrusions have obtained high-level permissions of enterprise cloud systems in less than 3 hours after locking the target. Taiwan's e-commerce SaaS platform provider 91APP has been attacked by ARTEX since October, which may affect up to 1,329 brands and about 4.8 million consumers, and some personal data may be leaked. The Cyberbiz e-commerce platform and a number of listed companies have also confirmed that they were hacked or had data leaked. The Ministry of Digital Development has issued an information security warning, reminding the public to beware of personal information fraud and calling on companies to check and fix vulnerabilities as soon as possible.

SpaceX launched the Starship giant rocket in Texas, challenging the "double capture" test of fully recovering the first and second stages for the first time. After liftoff, the booster malfunctioned, initiated the automatic abort procedure, crashed into the Gulf of Mexico and exploded, declaring a setback at the first hurdle; however, the spacecraft still successfully advanced towards orbit and strived to return on Saturday to complete mid-air clamping. The mission will deploy 26 Starlink satellites and orbit the earth 15 times at an altitude of approximately 275 kilometers.

Former OpenAI research engineer Will DePue claimed to have discovered NVIDIA Blackwell GPU hardware fingerprints from six model weights of DeepSeek, Dark Side of the Moon, MiniMax and Ant Group, raising concerns about Chinese AI companies' acquisition of high-end chips subject to U.S. export controls. However, the research cannot prove the source of the chips or violate export controls.

Huida CEO Huang Renxun resigned on the 12th, shocking the global technology industry. Counterpoint Research pointed out that the short-term AI chip product route and supply chain cooperation are expected to remain stable, but the strategy of the new management team needs to be observed in the medium and long term. It is predicted that server AI ASIC shipments will surpass GPUs in 2027, and the AI computing market will develop in parallel towards GPUs and customized ASICs.

Lex Fridman, an American technology podcaster, posted on the