Welcome to the [AI Daily] column! This is your daily guide to exploring the world of artificial intelligence. Every day, we bring you hot topics in AI, focus on developers, and help you understand technology trends and discover innovative AI product applications.
Fresh AI products Click to learn more: https://app.aibase.com/zh
1. Baidu upgrades AI Search and Search AI API, with call prices as low as RMB 4.9 per 1,000 calls
Baidu is extending AI Search from a Q&A tool to task delivery and real-time interaction, covering more than 170 scenarios, with GUI data scale increased 5x, showing its intent to build a service ecosystem; the API costs as low as RMB 4.9 per 1,000 calls, lowering the barrier for developers and expected to accelerate B2B adoption.
[AiBase Highlights:]
💰 Search AI API promotional price as low as RMB 4.9 per 1,000 calls, reducing developer integration costs.
🤖 Wenxin Assistant shifts from one-off Q&A to a deliverable, covering more than 170 application scenarios.
🎥 AI Lens adds dynamic tracking and supports real-time conversation in 9 dialects.
2. Doubao Work officially integrates DeepSeek, Kimi, and GLM, making enterprise procurement completely worry-free
Doubao Work aggregates flagship models such as Kimi, GLM, and DeepSeek. Enterprises do not need to purchase them separately; they can call them with AI points and switch flexibly. Administrators can control member, department, and terminal permissions with fine granularity, balancing efficiency and data security, highlighting the platformization trend of enterprise-grade AI assistants.
[AiBase Highlights:]
🧩 Doubao Work integrates three flagship large models: Kimi, GLM, and DeepSeek
💳 Enterprises do not need additional procurement; calls to third-party models directly deduct AI points
🔐 Administrators can finely control model usage permissions by member, department, and terminal
3. Edit documents directly in the chat window! Google Gemini upgrade: full support for inline editing of PDFs, spreadsheets, and code files
Google Gemini embeds document editing into the chat interface, completing the loop from generation to modification, covering formats such as PDFs, spreadsheets, and code, and reducing operational friction. Compared with relying on the Workspace sidebar, the native experience is stronger, and it also responds to competitors such as Claude. However, mobile support is limited, and collaboration on complex formats remains to be verified.
[AiBase Highlights:]
📄 PDF inline editing is supported on both web and mobile
🔄 Completes the post-generation modification loop without jumping to external apps
💬 You can select local content to ask questions or instruct conversion into charts
4. Moonshot AI announces that Kimi is officially integrated with Google Gmail, Drive, and Calendar
Moonshot AI's Kimi integrates with three major Google office services, strengthening productivity scenarios, ensuring security through OAuth, and improving the work efficiency of overseas users.
[AiBase Highlights:]
📧 Kimi integrates three major office services: Gmail, Google Drive, and Calendar
🔌 Users can set it up easily via the plugin page to summarize emails and plan schedules
🔒 All interactions are authorized via OAuth, ensuring account security and privacy
5. Check in for 30 days to get 8,800 credits for free: Doubao opens a 30-day free subscription to new and existing users
Doubao uses free subscription experiences and check-in credits to acquire and reactivate users, and its plan credit rules are more transparent; however, gifted credits are valid for 30 days and are restricted for video generation and Pro models, so long-term retention and paid conversion remain to be seen.
[AiBase Highlights:]
🎁 New and existing users can get a 30-day free subscription experience, and check in daily to receive 200 credits.
📊 The credits for the three tiers are clear every 7 days: 2,100, 5,600, and 14,000.
⏳ Checking in for 30 days earns up to 8,800 credits, valid for 30 days, with some advanced capabilities restricted.
6. AI can also rearrange and cover songs? Spotify plans to launch a paid music remix service
Spotify is shifting the focus of generative AI from "creating new songs" to "remixing licensed catalogs," reducing copyright risks and activating subscription monetization; plays are counted toward the original song's chart position and shared in revenue, which is expected to balance the interests of the platform, labels, and artists, but pricing and licensing negotiations remain key to implementation.
[AiBase Highlights:]
🤖 Spotify plans to launch AI music remixing in Q1 2027, with monthly fees possibly increasing by $4 to $5.
🎚️ Users can rearrange licensed songs or simulate singer covers, freely switching musical styles.
📜 It has reached licensing deals with Universal, Merlin, and others, and is still negotiating with Sony and Warner; plays count toward the original song and are shared in revenue.
7. Neuralink brain-computer interface gets its strongest upgrade: Grok connects to the brain, allowing patients to wake AI directly with their thoughts
This advance expands brain-computer interfaces from motor decoding to collaboration with large models. Through neural speech detection, closed-loop feedback, and system-level integration, it significantly reduces the patient's operational burden and validates the feasible path of "intent output + AI-managed execution." However, scale-up and long-term stability still require more clinical data.
[AiBase Highlights:]
🗣️ Neural speech detection enables zero-button thought-to-speech
🤖 Grok connects to the brain, and thoughts drive AI to perform tasks
📱 Lock-screen input and thought-based calls on phones become reality
8. Nvidia releases Nemotron 3.5 Transcribe speech transcription large model, accurately recognizing even when 8 people speak at once
Nvidia open-sources Nemotron 3.5 Transcribe. It uses a MoE architecture with 34 billion total parameters / 4.3 billion activated, supports transcription in 40 languages, translation between 31 languages, and separation of 8 speakers. It leads in WER on evaluations such as Open ASR, providing a high-precision foundation for the speech technology ecosystem.
[AiBase Highlights:]
🧠 MoE architecture, 34 billion total parameters, 4.3 billion activated parameters
🌍 Supports transcription in 40 languages and translation between 31 languages, plus separation of 8 speakers
🏆 5.33% word error rate on Open ASR, ranking first among open-weight models
