Total 671 · Clear filters
On 28 November 2023, at re:Invent, AWS introduced Amazon Q, a generative AI assistant for work, in preview. Answers respect users' permissions and cite sources, and Amazon Q Business became generally available on 30 April 2024.
On 21 November 2023 Anthropic released Claude 2.1, doubling the context window to 200,000 tokens. It also added a beta tool use feature and system prompts for setting Claude's role and style.
On 21 November 2023 Stability AI released Stable Video Diffusion, its first generative video foundation model, as a research preview of two image-to-video models with 14 and 25 frames.
On 13 November 2023 NVIDIA announced the HGX H200 at the SC23 conference in Denver, an addition to its Hopper line that twelve server partners and four major cloud providers were named to offer, with shipments expected from the second quarter of 2024.
On 6 November 2023 OpenAI held its first developer conference, DevDay, announcing GPT-4 Turbo with a 128K context window, custom GPTs built with a GPT Builder, and lower prices across its platform.
On 3 November 2023 xAI announced Grok, its first chatbot, powered by the new Grok-1 model, and offered early access to a limited number of US users.
On 2 November 2023 DeepSeek released its first model, DeepSeek Coder, a set of open-source code models from 1.3B to 33B parameters trained on 2 trillion tokens each.
On 2 November 2023, 01.AI released the open Yi-6B and Yi-34B base models. Yi-34B ranked first among open pre-trained models on two public leaderboards, ahead of Meta's Llama 2.
On 24 October 2023 iFlytek released Spark V3.0 and claimed it beat GPT-3.5 in Chinese across six areas, unveiled a Huawei-built computing platform the same day, and saw its stock fall to the daily limit.
On 17 October 2023 Adept released Fuyu-8B, an encoder-free multimodal model whose weights and example code went to Hugging Face the same day, positioned as part of a copilot for knowledge workers.
On 17 October 2023 Baidu CEO Robin Li unveiled ERNIE 4.0 at Baidu World 2023 in Beijing, claiming it was "not inferior in any aspect to GPT-4" and announcing a paid ERNIE Bot plan built on it.
In October 2023 Moonshot AI released Kimi, a chatbot built around very long context, at first in closed beta; it opened to the public in November 2023, and by March 2024 it could take in two million Chinese characters at once.
On 27 September 2023 Mistral AI released Mistral 7B, a 7.3-billion-parameter model with Apache 2.0 weights, less than five months after the company was founded. It was the start of the company's open-weight line.
At Meta Connect on 27 September 2023 Meta introduced Meta AI, a conversational assistant in US beta on WhatsApp, Messenger and Instagram, along with 28 AI characters, 14 of them played by celebrities and influencers. It was rebuilt on Llama 3 in April 2024.
On 25 September 2023, Preferred Networks published PLaMo-13B, its first large language model and its first open-weight release, trained on 1.5 trillion tokens and released under Apache-2.0.
On 20 September 2023 OpenAI announced DALL·E 3, its third-generation image model, which was then built directly into ChatGPT; it was designed to refuse requests that imitate the style of living artists.
On 13 September 2023 Stability AI released Stable Audio, a web tool that generates music and sound from text prompts, built on a latent diffusion model trained with licensed content from the AudioSparx library.
On 11 September 2023 Microsoft researchers posted the phi-1.5 technical report, a 1.3-billion-parameter model for natural-language common-sense reasoning, released openly under the MIT licence.
On 7 September 2023 Tencent unveiled Hunyuan, its proprietary large foundation model, at a summit in Shenzhen. It was presented to enterprises, with access through Tencent Cloud APIs.
On 7 September 2023 Anthropic introduced Claude Pro, its first paid consumer plan: $20 a month in the US or GBP 18 in the UK, with five times the free tier's usage of Claude 2, priority access at busy times and early access to new features.
On 6 September 2023 Baichuan released Baichuan 2, its next-generation open model family in 7B and 13B sizes, trained on 2.6 trillion tokens and published with intermediate checkpoints.
