每日AI简报（trustedreviews） · 2026/4/25

Anthropic 再获40亿美元融资，AI领域巨额资本竞赛加速

来源: https://www.theverge.com/2024/4/18/24135020/anthropic-amazon-google-funding-4-billion-forty-billion

标题

Anthropic just got $4 billion more from Amazon to fight the AI war

正文

Anthropic announced a significant new investment from Amazon on Thursday, with the cloud provider adding a fresh $2.75 billion to its stake in the AI company.

This brings Amazon's total invested amount in Anthropic to $4 billion, marking one of the largest corporate deals in the young, capital-intensive industry of building generative AI models.

In September, Anthropic announced that Amazon would invest up to $4 billion in the company and that the two would form a long-term partnership. Anthropic picked Amazon Web Services as its primary cloud provider and agreed to make its models available to AWS customers through Amazon Bedrock.

Amazon has now exercised its right to invest the remainder of that $4 billion, an Amazon spokesperson told The Verge. Anthropic said in an SEC filing that Amazon’s second investment was in the form of $2.75 billion in convertible notes. Previously, Amazon had invested $1.25 billion in Anthropic.

The news of the latest investment was first reported by Bloomberg.

Anthropic’s funding haul from Amazon follows $2 billion and $750 million in funding from Google last year. Google is also a key cloud and strategic partner for Anthropic, but Amazon was brought in as a “primary” cloud provider for mission-critical workloads.

Amazon and Google are not the only major tech companies pouring billions into Anthropic. Salesforce recently led a group of investors in contributing $450 million to the AI company, according to SEC filings, and the company has raised $7.6 billion in total.

The escalating investments from tech giants — and the fact that Amazon and Google, direct competitors, are both major backers — underscores the fierce battle for supremacy in generative AI and the high cost of developing state-of-the-art models.

Anthropic is the maker of Claude, one of the leading chatbots competing with OpenAI’s ChatGPT. The company has said it would use Amazon’s funding to advance its AI model development and invest in building better infrastructure.

This latest investment comes as the AI industry enters a critical phase, with companies racing to build more powerful models and find sustainable business models. While venture capital funding for AI startups remains hot, the scale of investment required for foundation model companies like Anthropic is on another level, often necessitating deep partnerships with cash-rich tech incumbents.

Meta开放Llama 3商用，打响“免费开源”AI模型之战

来源: https://www.theverge.com/2024/4/18/24135514/meta-llama-3-open-source-model-available-ai

标题

Meta’s Llama 3 is now available — and it’s open for commercial use

正文

Meta is opening Llama 3, its latest and most advanced large language model (LLM), to the public today alongside a new “real-time” image generator.

The company says Llama 3 is more capable than its predecessor, Llama 2, and that it will release multiple versions of the model in the coming months.

The two smaller versions being released today — an 8 billion and a 70 billion parameter model — will be available for developers to download for free. Meta says this is a “significant leap” from Llama 2, claiming the new model outperforms similarly sized competitors like Mistral 7B and Google’s Gemma.

“We believe these are the best open source models of their class, period,” Meta CEO Mark Zuckerberg said in a video announcing the launch.

Crucially, Llama 3 is being released with a commercial license, meaning developers and companies can use it freely with few restrictions, even for building commercial products. This is consistent with Meta’s approach with Llama 2.

The company is also launching an updated version of Meta AI, its in-house AI assistant, which will now be powered by Llama 3. It’s being integrated into Meta’s suite of apps, including Facebook, Instagram, WhatsApp, and Messenger, and is available through a new standalone website at meta.ai.

The Llama 3 launch comes hot on the heels of new models from competitors like Anthropic and Google. Meta’s strategy has leaned heavily on the “open” approach, releasing powerful models for free to foster developer adoption and build an ecosystem, while competitors like OpenAI, Anthropic, and Google often keep their most advanced models proprietary or behind APIs.

It’s a major push by Meta to plant its flag in the AI landscape with an open alternative to closed models like GPT-4, Claude 3, and Gemini. However, even with its open license, Llama 3’s “openness” is somewhat qualified. For training the model, Meta used enormous amounts of public data, but has not disclosed the specific datasets or fully detailed the training process, which is common in the industry but at odds with some definitions of “open source.”

In addition to the language models, Meta is also rolling out a new image generation feature called “Imagine.” It’s initially available only in the U.S. and can generate images in real-time as you type a prompt. The company says this is a beta feature and that the images will come with an invisible watermark to indicate they are AI-generated.