AI Daily · Nov 10
Nov 9 06:00 – Nov 10 06:00 (Beijing time) · 21 items
Updated 11/10 08:01 (Beijing time)
Today's perks
New perks in this issue come first; if there are fewer than 3, we add perks you can still claim.
- NewAlibaba QwenAlibaba Cloud Token Plan quotas raised, Pro monthly fee cut to ¥312Alibaba Cloud Token Plan personal edition quotas raised across all 4 tiers; Lite/Essential/Standard monthly fees unchanged, Pro monthly fee cut from ¥499 to ¥312, effective for existing subscribers as well.
- NewTencent Hunyuan / DeepSeekTencent WorkBuddy apologizes: DeepSeek-V4.1 Flash credits refunded at 50% offTencent WorkBuddy has apologized for not promptly announcing adjustments to its promotional campaign. Accounts that used DeepSeek-V4.1 Flash between Nov 5, 14:00 and Nov 8, 23:59 will be billed for credits at 50% off, with the 50% difference automatically refunded within 7 business days.
Today's highlights
- AnthropicAnthropic releases Claude Fable 5.2 and Mythos 5.2 modelsAnthropic launches Claude Fable 5.2 and Claude Mythos 5.2, claiming improvements in coding, reasoning, and vision, with output speeds up to 30% faster than the previous generation and API pricing reduced by 20%.
- CursorClaude Fable 5.2 launches on Cursor, scores up to 60% on CursorBenchCursor officially announced that Claude Fable 5.2 is now available in Cursor. The model scores 60% on CursorBench at the highest effort level and 56.9% at the default level, higher than Fable 5.1's 49.2%.
- OpenCodeMeta releases lightweight model Muse Ember, integrated into OpenCodeMeta releases Muse Ember, a new lightweight model for agentic execution, at 10 cents per million input tokens, now available in OpenCode.
Perks and offers
Alibaba QwenThird-party report
Alibaba Cloud Token Plan quotas raised, Pro monthly fee cut to ¥312
Alibaba Cloud Token Plan personal edition quotas raised across all 4 tiers; Lite/Essential/Standard monthly fees unchanged, Pro monthly fee cut from ¥499 to ¥312, effective for existing subscribers as well.
Tencent Hunyuan
DeepSeekEndedThird-party report
Tencent WorkBuddy apologizes: DeepSeek-V4.1 Flash credits refunded at 50% off
Tencent WorkBuddy has apologized for not promptly announcing adjustments to its promotional campaign. Accounts that used DeepSeek-V4.1 Flash between Nov 5, 14:00 and Nov 8, 23:59 will be billed for credits at 50% off, with the 50% difference automatically refunded within 7 business days.
Models and products
Anthropic
Anthropic releases Claude Fable 5.2 and Mythos 5.2 models
Anthropic launches Claude Fable 5.2 and Claude Mythos 5.2, claiming improvements in coding, reasoning, and vision, with output speeds up to 30% faster than the previous generation and API pricing reduced by 20%.
- Cursor
Claude Fable 5.2 launches on Cursor, scores up to 60% on CursorBench
Cursor officially announced that Claude Fable 5.2 is now available in Cursor. The model scores 60% on CursorBench at the highest effort level and 56.9% at the default level, higher than Fable 5.1's 49.2%.
OpenCode
Meta releases lightweight model Muse Ember, integrated into OpenCode
Meta releases Muse Ember, a new lightweight model for agentic execution, at 10 cents per million input tokens, now available in OpenCode.
Anthropic
Anthropic releases Claude model selection guide
Anthropic officially released a new guide explaining when to choose Fable 5.2, Opus 5.5, Sonnet, or Haiku, and introducing when to adjust effort instead of switching models, as well as how to run subagents with smaller models to save costs.
Google AI
NVIDIA builds robot demo on Gemma 4 E4B
Google Gemma's official account showcased an NVIDIA robot demo built on Gemma 4 E4B: the model runs on an 8GB Jetson with integrated real-time conversation, tool calling, animation, and vision models, and can scan a room and play a movie.
Google AI
Google Maps Platform integrates with AI Studio
Google announced that developers can now build applications using Google Maps Platform in AI Studio, with one-click setup, map app generation from descriptions, and access to map tiles, places, weather, routes, and 3D capabilities.
Mistral AI
Mistral uses Voxtral Transcribe 3 to transcribe AI Engineer in real time…
Mistral says it used Voxtral Transcribe 3 to transcribe every main-stage talk at AI Engineer Paris in real time, feeding in speaker names and conference technical terms through contextual biasing to improve accuracy.
OpenAI
OpenAI allows multiple local computers to connect to the same dot
OpenAI developer accounts announced that users can now connect multiple local computers to the same dot and choose which one executes tasks, for example building and testing an app on a dedicated machine while continuing to write code on a laptop.
OpenAI
OpenAI Codex launches ModRetro Chromatic plugin
The OpenAI developer account announced that Codex ModRetro Chromatic is now available on Supply Co. With the ModRetro Chromatic plugin in Codex, users can design levels, create sprite animations, and playtest them on the handheld.
xAI
Musk: Grok Bot team fixed issues overnight, experience will improve noticeably tomorrow
Musk said he held a roundtable retrospective with the Grok Bot engineering team; fixes will roll out overnight, and users should feel a noticeable improvement the next day; if issues remain, reply under the post with feedback.
Media reports
- Related companies
Alibaba QwenThird-party report
Qwen Office 2.0 released: the first to support multi-user and Agent collaboration
On Nov 9, Qwen Office 2.0 was officially released, introducing the first "Collaboration" feature that supports multi-user and Agent collaboration, and launching core capabilities such as "Human-AI Co-creation", "Material Space", "Automated Tasks", and "Three-Device Interconnection". All features are now fully available in the Qwen Office client.
- Related companies
Google AIThird-party report
Google releases 2 new AI chips: TPU 8tx and TPU 8i Max
Google expands its AI chip lineup with the TPU 8tx and TPU 8i Max, both built on the 8th-generation TPU architecture and respectively targeting secure AI and ultra-long-context model inference needs.
- Related companies
Tencent HunyuanThird-party report
Tencent Hunyuan Hy4 released, tested with WorkBuddy multimodal Agent
QbitAI reports that Tencent Hunyuan's new model Hy4 ranks among the strongest open-source models, and tested its delivery performance on multimodal Agent tasks when paired with WorkBuddy.
- Related companies
Google AIThird-party report
US senator writes to Google, asks it to let schools disable search AI features
US Senator Markey sent a letter to Google asking it to allow schools to disable the artificial intelligence features in its search engine. The letter cited reports that Google plans to disable students' access to Gemini and search AI services by default, leaving it to schools to decide whether to re-enable them.
- Related companies
MetaThird-party report
Meta expands AI age estimation globally to strengthen teen account protections
Meta announced the global rollout of AI age estimation technology, which will place teens in Instagram teen account protections even if they enter an adult birth date. The feature covers Instagram, Facebook, and Messenger, and Meta faces multiple lawsuits alleging its platforms harm underage users.
- Related companies
Google AIThird-party report
Google Firebase Studio to shut down in March 2027
According to Engadget, Google's online IDE Firebase Studio will officially shut down on Mar 22, 2027, with data permanently deleted, and users need to move to alternatives such as Google Antigravity or Google AI Studio.
- Related companies
OpenAIThird-party report
Study: ChatGPT accounts for over 81% of global AI chatbot traffic
Graphite research shows that ChatGPT accounts for over 81% of total traffic among major AI chatbots worldwide, with no competitor reaching a 10% share; its monthly visits hit a record high of 61.35 billion in September.
- Related companies
AnthropicThird-party report
Anthropic study: scientists mostly use Claude for data analysis and writing
Anthropic released a research report based on 207,000 scientific research sessions from Aug 2 to Sep 30, 2026, showing that scientists mainly use Claude for data analysis and writing, while AI is used less for collecting data and generating research hypotheses. Usage is more frequent in computer science, mathematics, and statistics.
- Related companies
Baidu ERNIEThird-party report
Baidu ERNIE discloses RSI progress for the first time, ERNIE Evolve already used in model R&D
On Nov 9, Baidu ERNIE disclosed its latest RSI progress for the first time, stating that ERNIE Evolve has already been used in the R&D of its latest models, and ERNIE 5.5 Flash may be released soon.
Generates a long image you can save and share.