Runway releases Gen-4
On 31 March 2025 Runway released Gen-4, a video generation model built to keep characters, objects and locations consistent across shots, controlled with visual references and text instructions rather than fine-tuning.
Total 1,098
CoreWeave goes public on the Nasdaq at $40 a share
CoreWeave listed on the Nasdaq on 28 March 2025 at $40 a share, raising $1.5 billion in the biggest US tech IPO since 2021, with NVIDIA anchoring the deal.
xAI acquires X Corp. in an all-stock deal
On 28 March 2025 Elon Musk announced that his AI company xAI had acquired X Corp., the owner of the X social network, in an all-stock deal that valued X at $33 billion and xAI at $80 billion.
Alibaba releases Qwen2.5-Omni, a 7B model that sees, hears and talks
On 26 March 2025 Alibaba released Qwen2.5-Omni-7B, an end-to-end model that perceives text, images, audio and video and answers with text and natural speech.
Google releases Gemini 2.5 Pro, its first "thinking model"
On 25 March 2025 Google announced Gemini 2.5 and introduced Gemini 2.5 Pro Experimental, a thinking model that debuted at #1 on LMArena and became generally available on 17 June 2025.
ChatGPT gets native GPT-4o image generation
On 25 March 2025 OpenAI added native image generation to GPT-4o in ChatGPT, replacing DALL·E 3 as the default option; the same model reached developers a month later as gpt-image-1.
DeepSeek updates V3 as DeepSeek-V3-0324 under the MIT License
In March 2025 DeepSeek updated DeepSeek-V3 as DeepSeek-V3-0324 and moved the model to the MIT License, the same license used for DeepSeek-R1.
Tencent officially releases the Hunyuan-T1 reasoning model
On 21 March 2025 Tencent launched Hunyuan-T1, a deep-reasoning model offered on Tencent Cloud and priced against DeepSeek-R1.
Claude gains web search
On 20 March 2025 Anthropic added web search to Claude, letting the model decide when to look things up and cite sources; it began as a preview for paid US users and became available globally on all plans on 27 May 2025.
METR publishes the 'time horizon' metric for AI task completion
A March 2025 paper led by Thomas Kwa introduced the "50%-task-completion time horizon", measuring how long a task takes humans before AI can finish it half the time; frontier models reached about 50 minutes.
NVIDIA announces Blackwell Ultra
NVIDIA announced Blackwell Ultra at GTC in San Jose on 18 March 2025, a platform aimed at AI training and at inference that spends more compute per answer, alongside new HGX B300 NVL16 systems and named server and cloud partners.
NVIDIA announces Isaac GR00T N1
On 18 March 2025 NVIDIA announced Isaac GR00T N1, a humanoid robot foundation model released immediately to robotics developers as the first of a family of customizable pretrained models, alongside open simulation tools.
NVIDIA launches the Llama Nemotron open reasoning models
On 18 March 2025, NVIDIA launched the Llama Nemotron family of open reasoning models at its GTC conference, with partners including Microsoft and SAP.
01.AI launches the Wanzhi enterprise LLM platform
On 17 March 2025, Chinese AI company 01.AI released Wanzhi, an enterprise one-stop platform for large models that supports DeepSeek, Qwen and its own Yi models, alongside customized DeepSeek deployment and fine-tuning tools.
Mistral AI releases Mistral Small 3.1 with image input
On 17 March 2025 Mistral AI released Mistral Small 3.1, an Apache 2.0 model that adds image understanding and a 128K context window while running on a single RTX 4090 or a 32GB Mac.
Baidu makes ERNIE Bot free ahead of schedule
On 16 March 2025 Baidu made its ERNIE Bot free for individual users, two weeks ahead of its planned 1 April date, giving free access to the ERNIE 4.5 and ERNIE X1 models.
Baidu launches ERNIE X1, its first reasoning model
On 16 March 2025 Baidu launched ERNIE X1, a deep-thinking reasoning model and its first multimodal reasoning model able to use tools. Baidu said it matched DeepSeek-R1 at half the price.
Baidu releases ERNIE 4.5, a native multimodal foundation model
On 16 March 2025 Baidu released ERNIE 4.5, a native multimodal foundation model, together with the reasoning model ERNIE X1, and made ERNIE Bot free to individual users ahead of schedule.
Ai2 releases OLMo 2 32B
On 13 March 2025, Ai2 released OLMo 2 32B, which it called the first fully open model to beat GPT-3.5-Turbo and GPT-4o mini on a suite of academic benchmarks, and published everything needed to reproduce it.
Cohere launches Command A, an enterprise model that runs on two GPUs
On 13 March 2025 Cohere introduced Command A, a 111-billion-parameter model that runs on two GPUs. Cohere claimed it matches or beats GPT-4o and DeepSeek-V3 on agentic enterprise tasks.
Google releases Gemma 3
On 12 March 2025 Google released Gemma 3, a family of four model sizes from 1B to 27B with image input and a 128,000-token context window, alongside ShieldGemma 2, an image-safety checker.