News & Updates
June 9, 2026
Abogen releases v1.3.1: new Web UI, EPUB 3 pipeline and Supertonic TTS engine
Abogen v1.3.1 brings a Flask-based web UI, EPUB 3 media overlay packaging pipeline, Supertonic TTS engine, multitimbral role assignment, and Caliber/Audiobookshelf integration.
Source: Abogen GitHub Releases
Anthropic Releases Claude Fable 5 and Mythos 5 – Strongest Models Yet, Mythos Level with Safety Guardrails
Anthropic launches Mythos level models Fable 5 (universal version with safety guardrail) and Mythos 5 (cyber defense version). Stripe can complete two months' worth of work in one day, speeding up drug design by 10 times, and is priced at $10/$50 per million tokens.
Source: Anthropic official announcement
June 11, 2026
You.com strengthens multi-interface Web search capabilities, and the Agent networking standard layer continues to expand
You.com Improves enterprise Agent networking capabilities through the combination of Search, Contents, Research, and Finance Research APIs.
Source: You.com official website and Pricing
Tavily accelerates the implementation of Agent network search: from "can be searched" to "can be connected to production"
Tavily has exposed more complete search, extraction and research interface capabilities, and is advancing Agent network retrieval from the experimental stage to the infrastructure layer that can be produced and accessed.
Source: Tavily official website, Tavily Docs, Tavily Pricing
Sora promotes parallelization of content production and moves the creative review link forward
Sora The core influence on the content team is not to simply increase the speed of video generation, but to move creative screening forward to the parallel generation stage.
Source: OpenAI official release, Sora product page, creative tool platform observation, content production industry analysis
SerpApi strengthens the AI search interface matrix, and the search data infrastructure enters the "Agent ready" stage
SerpApi continues to expand AI search-related interfaces based on the original SERP structured capabilities, further consolidating its positioning as the underlying capability layer for Agent networked retrieval.
Source: SerpApi official website, Pricing page, API list page
Perplexity launches Comet: AI browser begins to push search to the task execution layer
The Comet launched by Perplexity is not just a search portal, but attempts to extend information retrieval into task execution and redefine the evaluation criteria of AI browsers.
Source: Perplexity official product page, official demo, Hacker News, Product Hunt and other third-party media
n8n strengthens AI orchestration capabilities, and workflow testing and governance requirements are simultaneously increased
n8n is moving from automation connectors to AI orchestration layers. While enterprises gain stronger process capabilities, they also need to simultaneously improve testing, monitoring, and rollback mechanisms.
Source: n8n official product page, official documentation, GitHub discussion, industry media reports
Secret Tower AI search continues to strengthen in-depth research and fact-checking, and the Chinese search experience continues to upgrade
Metaso AI Search Continuously optimize retrieval, research and voice capabilities around Chinese information acquisition scenarios.
Source: Secret Tower official website
Manus Agent platform enters the field of enterprise evaluation, and governance base becomes a prerequisite
When teams evaluate agent platforms like Manus , the first thing they need to confirm is the governance boundary, not the demonstration effect.
Source: Manus official product page, enterprise automation cases, AI Agent platform review
Linkup clarifies the Fetch/Search/Research cost model, making it easier to implement enterprise AI search platforms
Linkup Provides clear billing intervals in addition to performance signals to enhance enterprise deployment predictability.
Source: Linkup official website and Pricing
DeepSeek API cache billing system failure, proactively returning bonuses to affected users
Some errors occurred in the DeepSeek API cache billing system between May 13 and June 9. The team took the initiative to return the corresponding amount to the affected users in the form of bonuses, reflecting its emphasis on the developer ecosystem.
Source: IT Home
Dappier improves the price gradient of AI Answers and makes the commercialization path of media AI clearer
Dappier Discloses the layering of solutions from Free to Enterprise, strengthening the integrated positioning of media AI Q&A and monetization.
Source: Dappier official website and Pricing
CometAPI: One Key can access 500+ models, and the average price is 20-40% lower than the original price.
CometAPI provides a unified API gateway that is compatible with OpenAI SDK and is connected to 500+ mainstream models. The overall pricing is 20-40% lower than that of the original model, and there is no monthly fee based on volume.
Source: CometAPI Official
ChatGPT enterprise capabilities continue to expand, and process handover becomes the key to implementation
Enterprise AI projects are difficult to scale, often not because of insufficient model capabilities, but because the results cannot smoothly enter process nodes that can be handed over, accountable, and reusable.
Source: OpenAI official product release, ChatGPT official documentation, enterprise application cases, AI collaboration tool evaluation
June 18, 2026
GPT-5.5 Instant health intelligence breakthrough: 230 million weekly active users, doctors rated it as super human
GPT-5.5 Instant reaches the level of cutting-edge models in health assessment, with doctors evaluating its answers to exceed those of human doctors. The factual question rate dropped by 71% in two months, and it is open to all free users.
Source: OpenAI Official Blog
May 22, 2026
DeepSeek V4-Pro API permanently reduced to 1/4 of the original price, cache hit only 0.025 yuan/million tokens
The price of DeepSeek V4-Pro API has been permanently reduced to 1/4 of the original price, and the cache hit is only 0.025 yuan/million tokens. The 25% discount will be permanent, and the pricing strategy is extremely aggressive.
Source: IT Home
May 28, 2026
DeepSeek service "partial outage" was repaired half an hour later, with a total of five failures in May
The DeepSeek service was partially interrupted on the morning of May 28 and was repaired about half an hour later. There have been five service problems this month, two of which were complete interruptions.
Source: IT Home
May 23, 2026
DeepSeek API output speed-up: supports 500 concurrency by default, enterprises can apply for larger scale
DeepSeek announced that the API has completed output speedup and service expansion. By default, it supports 500 concurrency online at the same time. Enterprise users can apply for a larger concurrency scale online. It will be promoted simultaneously with the V4-Pro price reduction strategy.
Source: IT Home
June 3, 2026
Big model price war: Xiaomi follows DeepSeek and cuts prices like crazy, Luo Fuli advises peers not to blindly follow suit
After DeepSeek announced that the V4-Pro discount would be permanent, Xiaomi quickly followed up with a discount of up to 99%. Luo Fuli reminded her peers not to blindly follow the trend. The large model industry went from "price increase tide" to "price reduction war" in just a few weeks.
Source: 36Kr (Rhodium Technology)