# forck.live — 16–22 September 2024 > AI lab announcements, news reports and podcasts. forck.live follows what companies announce about their own products. Press may corroborate an official story; it never leads and never appears as an item of its own. A missing date stays missing. A title is the publisher's words — where a company does not publish in English, an English rendering is listed separately as `Title in English` and named as this site's translation, never as theirs. This file is one issue of the weekly archive: every confirmed first-party announcement of ISO week 2024-W38, 16–22 September 2024, newest first. The window never moves, so this URL always names the same seven days — but a post published inside them and detected later is added when it arrives, so a closed week can still gain entries. Every entry is one first-party announcement, and carries the same eight fields in this order before any others: Company · Announced · Category · Coverage · Announcement · Group · Source · Record `Source` is the company's own page the announcement was traced to and `Record` is this site's page for it. Both are safe to cite. `Coverage` is a number of outlets, or `not counted` — which means this record reached no count of its own, never that nobody wrote about it. `Category` is this tracker's own analysis, and is authoritative wherever it is stated. `not stated` means either that nothing has analysed the post yet or that the analysis placed it outside this taxonomy; neither is a verdict about the announcement. `Announcement` is whether the post is news of something the company did, rather than a how-to, a customer story, a changelog listing or a staffing note. `Group` is which of this record's groups the entry falls in, carried on the entry itself so it holds where the groups are headings and where they are not. One of: `models`, `covered`, `announcements`, `routine`. Anything after `Record` is held detail that only some announcements carry. - This week's page: https://forck.live/week/2024-W38 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Vercel — From CDNs to Frontend Clouds - Company: Vercel (vercel.com) - Announced: 2024-09-20T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/from-cdns-to-frontend-clouds - Record: https://forck.live/items/20948-from-cdns-to-frontend-clouds - Subject: v0 / AI SDK / AI Gateway How UX demands and DIY infrastructure headaches shaped the next generation of web application delivery Web apps today are judged on six core principles: speed , global presence , scalability , dynamism , reliability , and security . ### Vercel — AI SDK 3.4 - Company: Vercel (vercel.com) - Announced: 2024-09-20T13:00:00+00:00 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/ai-sdk-3-4 - Record: https://forck.live/items/20947-ai-sdk-3-4 - Subject: v0 / AI SDK / AI Gateway Vercel released AI SDK 3.4, adding language model middleware for intercepting and modifying model calls, a data stream protocol enabling AI SDK UI to work with any backend, structured output modes, multi-step tool use, tracing improvements, and mock models for testing. ### Replicate — Using synthetic training data to improve Flux finetunes - Company: Replicate (replicate.com) - Announced: 2024-09-20 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/using-synthetic-data-to-improve-flux-finetunes - Record: https://forck.live/items/2513-using-synthetic-training-data-to-improve-flux-finetunes - Subject: Platform Three techniques for using synthetic training data to improve Flux fine-tuned models: generating training data from a single image using the consistent-character model, using outputs from the fine-tuned model itself as training data, and combining multiple LoRA styles to diversify training data. ### Hugging Face — Optimize and deploy with Optimum-Intel and OpenVINO GenAI - Company: Hugging Face (huggingface.co) - Announced: 2024-09-20 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/deploy-with-openvino - Record: https://forck.live/items/1828-optimize-and-deploy-with-optimum-intel-and-openvino-genai - Subject: Platform Optimizing and deploying Hugging Face Transformers models using Optimum-Intel and OpenVINO GenAI, focusing on edge and client-side deployment with C++ and Python. It covers environment setup, exporting models to OpenVINO IR, weight-only quantization (including INT4/INT8 with AWQ and scale estimation), and deployment using the OpenVINO GenAI API. ### OpenAI — Genmab launches “AI Everywhere” - Company: OpenAI (openai.com) - Announced: 2024-09-19T04:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/genmab - Record: https://forck.live/items/706-genmab-launches-ai-everywhere - Subject: GPT / ChatGPT / API Genmab adopts ChatGPT Enterprise, with support from OpenAI's security and privacy commitment. ### Alibaba — Qwen2.5: A Party of Foundation Models! - Company: Alibaba (alibaba.com) - Announced: 2024-09-18T16:00:04+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwen2.5/ - Record: https://forck.live/items/2305-qwen2-5-a-party-of-foundation-models - Subject: Qwen Alibaba announces Qwen2.5, a new generation of foundation models, as an open-source release. ### Alibaba — Qwen2.5-LLM: Extending the boundary of LLMs - Company: Alibaba (alibaba.com) - Announced: 2024-09-18T16:00:03+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwen2.5-llm/ - Record: https://forck.live/items/2306-qwen2-5-llm-extending-the-boundary-of-llms - Subject: Qwen Alibaba's Qwen team announces the Qwen2.5 series of decoder-only dense language models, with seven open-sourced models ranging from 0.5B to 72B parameters, and plans to open-source Qwen2. ### Alibaba — Qwen2.5-Coder: Code More, Learn More! - Company: Alibaba (alibaba.com) - Announced: 2024-09-18T16:00:02+00:00 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwen2.5-coder/ - Record: https://forck.live/items/2307-qwen2-5-coder-code-more-learn-more - Subject: Qwen Alibaba releases Qwen2.5-Coder, the next generation of open-source coding models, and renames CodeQwen to Qwen-Coder. ### Alibaba — Qwen2.5-Math: The world's leading open-sourced mathematical LLMs - Company: Alibaba (alibaba.com) - Announced: 2024-09-18T16:00:01+00:00 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwen2.5-math/ - Record: https://forck.live/items/2308-qwen2-5-math-the-world-s-leading-open-sourced-mathematical-llms - Subject: Qwen Alibaba's Qwen team announces the open-source release of the Qwen2.5-Math series, including base models (1.5B, 7B, 72B), instruction-tuned variants, and a mathematical reward model, designed for solving English and Chinese math problems through chain-of-thought and tool-integrated reasoning. ### Runway — Runway Partners with Lionsgate - Company: Runway (runway.com) - Announced: 2024-09-18T10:10:10.110000+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://runway.com/news/runway-partners-with-lionsgate - Record: https://forck.live/items/7923-runway-partners-with-lionsgate - Subject: Gen / Aleph Runway and Lionsgate have entered into a partnership to create and train a new AI model customized on Lionsgate's proprietary catalog, designed to help Lionsgate Studios and its creative talent augment their work by generating cinematic video that can be iterated using Runway's tools. ### Kyutai — Moshi open-source release: run Moshi locally! - Company: Kyutai (kyutai.org) - Announced: 2024-09-18 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://kyutai.org/blog/2024-09-18-moshi-release - Record: https://forck.live/items/18360-moshi-open-source-release-run-moshi-locally - Subject: Moshi / Pocket TTS ← Back to Blog Moshi open-source release: run Moshi locally! 18 September 2024 ### Hugging Face — Fine-tuning LLMs to 1.58bit: extreme quantization made easy - Company: Hugging Face (huggingface.co) - Announced: 2024-09-18 - Category: open-weight-release - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/1_58_llm_extreme_quantization - Record: https://forck.live/items/1829-fine-tuning-llms-to-1-58bit-extreme-quantization-made-easy - Subject: Platform Hugging Face fine-tuned Llama3 8B models to 1.58-bit quantization using the BitNet architecture, released the models under the HF1BitLLM organization, and introduced a new quantization method called 'bitnet' in Transformers. ### Vercel — Managing 275 thousand pages and 8 million assets at top speed with ISR - Company: Vercel (vercel.com) - Announced: 2024-09-17T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/managing-275-thousand-pages-and-8-million-assets-with-isr - Record: https://forck.live/items/20949-managing-275-thousand-pages-and-8-million-assets-at-top-speed-with-isr - Subject: v0 / AI SDK / AI Gateway Digital agency Americaneagle.com helped Mecum Auctions redesign their digital platform using a composable stack with Vercel's Frontend Cloud and Incremental Static Regeneration (ISR). The new architecture manages 275,000 lot pages and 8 million assets, generating pages on-demand as users request them, and allows the marketing team to create and update content without IT support. ### OpenAI — Using GPT-4 to improve teaching and learning in Brazil - Company: OpenAI (openai.com) - Announced: 2024-09-17T05:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/arco-education - Record: https://forck.live/items/707-using-gpt-4-to-improve-teaching-and-learning-in-brazil - Subject: GPT / ChatGPT / API OpenAI announces an initiative to improve teaching and learning in Brazil. ### Mistral AI — Announcing Pixtral 12B - Company: Mistral AI (mistral.ai) - Announced: 2024-09-17 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://mistral.ai/news/pixtral-12b - Record: https://forck.live/items/3342-announcing-pixtral-12b - Subject: Mistral / Le Chat Mistral AI announces Pixtral 12B. ### Mistral AI — AI in abundance - Company: Mistral AI (mistral.ai) - Announced: 2024-09-17 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://mistral.ai/news/september-24-release - Record: https://forck.live/items/3341-ai-in-abundance - Subject: Mistral / Le Chat The source text is a brief announcement titled 'AI in abundance' dated September 17, 2024, by Mistral AI team, but no further details are provided. ### Hugging Face — Introducing the SQL Console on Datasets - Company: Hugging Face (huggingface.co) - Announced: 2024-09-17 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/sql-console - Record: https://forck.live/items/1830-introducing-the-sql-console-on-datasets - Subject: Platform Hugging Face announced the SQL Console, a new feature on the Hugging Face Hub that allows users to run SQL queries directly on datasets using DuckDB WASM in the browser, with features like 100% local execution, full DuckDB syntax, export to Parquet, and shareable links. ### Replit — Introducing Replit Agent - Company: Replit (replit.com) - Announced: 2024-09-16T17:36:53+00:00 - Category: product-launch - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/introducing-replit-agent - Record: https://forck.live/items/17977-introducing-replit-agent - Subject: Replit Agent Replit launched Replit Agent, an AI system that creates and deploys applications from a few sentences. The agent configures the development environment, installs dependencies, and executes code. It is available in early access to all Replit Core subscribers and should be treated as alpha software. ### Harvey — Harvey to Expand Team with New London Office - Company: Harvey (harvey.ai) - Announced: 2024-09-16T15:10:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/harvey-to-expand-team-with-new-london-office - Record: https://forck.live/items/17726-harvey-to-expand-team-with-new-london-office - Subject: Harvey Harvey is opening a London office to serve as its European headquarters, supporting its growing EMEA customer base and expanding its sales, business development, customer success, and AI research teams. ### Vercel — ISR: A flexible way to cache dynamic content - Company: Vercel (vercel.com) - Announced: 2024-09-16T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/isr-a-flexible-way-to-cache-dynamic-content - Record: https://forck.live/items/20950-isr-a-flexible-way-to-cache-dynamic-content - Subject: v0 / AI SDK / AI Gateway Vercel's guide explains Incremental Static Regeneration (ISR), a caching strategy that combines static site generation and server-side rendering to serve dynamic content at cached speeds. It covers benefits such as reduced server load, selective regeneration, and on-demand invalidation, and provides examples across ecommerce, news, and AI applications. ### OpenAI — An update on our safety & security practices - Company: OpenAI (openai.com) - Announced: 2024-09-16T13:00:00+00:00 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/update-on-safety-and-security-practices - Record: https://forck.live/items/708-an-update-on-our-safety-security-practices - Subject: GPT / ChatGPT / API OpenAI provides an update on their safety and security practices. ### Runway — Introducing the Runway API for Gen-3 Alpha Turbo - Company: Runway (runway.com) - Announced: 2024-09-16T10:10:10.110000+00:00 - Category: api-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://runway.com/news/introducing-the-runway-api - Record: https://forck.live/items/7924-introducing-the-runway-api-for-gen-3-alpha-turbo - Subject: Gen / Aleph Runway announced the launch of its API for the Gen-3 Alpha Turbo model, providing developers with access for integration into applications and products. The API is initially available to select partners, with a wider release planned in the coming weeks. ### Hugging Face — Introducing Community Tools on HuggingChat - Company: Hugging Face (huggingface.co) - Announced: 2024-09-16 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/community-tools - Record: https://forck.live/items/1831-introducing-community-tools-on-huggingchat - Subject: Platform Hugging Face announces Community Tools on HuggingChat, a feature that allows users to turn any Hugging Face Space into a tool usable by models directly from HuggingChat, expanding modalities to include image understanding, video generation, and text-to-speech.