Total 1,098
On 5 September 2024 01.AI released the open Yi-Coder code models in 1.5B and 9B sizes. The 9B chat model scored 23.4% on LiveCodeBench, above every other model under 10 billion parameters, according to the company.
On 4 September 2024 Baidu upgraded and renamed its ERNIE Bot app as Wenxiaoyan (文小言), repositioning it as a "new search" assistant that combines search with text, image and multimodal generation.
In September 2024 Ai2 released OLMoE, a mixture-of-experts language model with 7B total but 1B active parameters per token, and published its weights, data, code and training logs.
In September 2024, ChatGPT Plus and Team subscribers got Advanced Voice Mode, a real-time spoken conversation feature built on GPT-4o. Its rollout, the removal of one of its launch voices, and its later replacement trace how voice became a default part of ChatGPT.
In September 2024 MiniMax released video-01, a video-generation model inside its Hailuo AI platform, which turns text prompts into short clips; the model later became the focus of a U.S. copyright suit filed by major studios.
In September 2024 Kuaishou released Kling 1.5, a new version of its AI video generator that added 1080p output, improved motion and prompt adherence, and a motion brush, while keeping the price unchanged.
In September 2024 Anthropic announced Claude for Enterprise, aimed at organisations that work with internal knowledge. It offered a 500K-token context window, single sign-on, audit logs and a GitHub integration.
xAI's Colossus supercomputer in Memphis, Tennessee, built with NVIDIA, went into operation in 2024 with about 100,000 H100 GPUs. It was the main training system behind Grok.
On 29 August 2024 Alibaba released Qwen2-VL, a vision-language model that understands videos over 20 minutes long. The 2B and 7B models were open-sourced under Apache 2.0, and the 72B model was initially API-only.
On 22 August 2024 AI21 Labs published Jamba 1.5, a pair of hybrid Transformer-Mamba models. It also open-sourced ExpertsInt8, a quantisation technique that lets the Large model run on eight 80 GB GPUs.
On 18 August 2024 AgiBot released five commercial humanoid robots in its Yuanzheng and Lingxi families, opened mass production, set out a G1–G5 embodied-intelligence roadmap, and launched open-source and ecosystem programmes.
On August 13, 2024, OpenAI released SWE-bench Verified, a 500-sample human-validated subset of SWE-bench built with the benchmark's original authors from Princeton, after filtering ambiguous tasks and unreliable tests.
On 13 August 2024 xAI released Grok-2 and Grok-2 mini in beta to X Premium subscribers. In August 2025 it published the Grok 2.5 weights on Hugging Face.
In August 2024 a US judge let illustrators suing Stability AI, Midjourney, DeviantArt and Runway AI pursue some copyright and trademark claims, while leaving the core training-data question undecided.
On 8 August 2024 Hugging Face announced it had acquired XetHub, a Seattle startup whose technology lets Git scale to terabyte-sized repositories by chunking and deduplicating data; Hugging Face planned to move the Hub off Git LFS onto a Xet-based backend.
Figure AI released Figure 02, a humanoid robot with redesigned hardware and software, six onboard cameras, an onboard vision-language model and hands that can grasp 25 kg, as it works toward mass production for industrial users.
On 6 August 2024 Zhipu AI and Tsinghua University opened the CogVideoX text-to-video models, starting with CogVideoX-2B. The 5B version followed on 27 August 2024, and the family is the open-source counterpart of Zhipu's Ying product.
The EU Artificial Intelligence Act took effect on 1 August 2024. It sets one risk-based rulebook for AI across EU countries, banning social scoring and requiring chatbots to disclose themselves, with rules for general-purpose models to follow.
On 1 August 2024 Black Forest Labs announced itself and released the FLUX.1 text-to-image models, backed by a $31 million seed round led by Andreessen Horowitz.
In August 2024 Google signed a non-exclusive deal to license Character.AI's large language model technology and hired co-founder and CEO Noam Shazeer, reportedly paying $2.7 billion; Shazeer returned to a technical lead role on Gemini.
In August 2024 Microsoft released three Phi-3.5 models, mini, MoE and vision, all with a 128K-token context and under the MIT licence.
