# Mistral AI 为 le Chat 推出重大更新（Beta）

- 来源：Mistral AI：News（网页）
- 发布时间：2024-11-18 00:00
- AIHOT 分数：61
- AIHOT 链接：https://aihot.virxact.com/items/cmppdcr7e0e45slv4bavwu7af
- 原文链接：https://mistral.ai/news/mistral-chat

## AI 摘要

Mistral AI 对其免费 AI 工作助手 le Chat 进行了重大更新（Beta）。新增功能包括：支持带引用的网络搜索；用于协作创作的 Canvas 界面；由新多模态模型 Pixtral Large 驱动的文档与图像理解功能；由 Black Forest Labs Flux Pro 支持的图像生成能力；以及更快的响应速度（由推测编辑技术驱动）。此次更新也展示了 le Chat 在提供前沿模型访问、网络搜索、Canvas 等多项核心功能方面的免费优势。所有新功能将陆续向用户开放。

## 正文

AI is as much a pursuit of seamless interaction between human and machine intelligence, as it is of frontier science that makes machines reason better. At Mistral AI, we began our journey by developing state of the art models; we have now begun to invest in AI interfaces, starting with some big updates to le Chat .

If you’re new to le Chat, welcome! Le Chat [/lə ʃa/, French: ‘the cat’] is the free generative AI work assistant from Mistral AI. You can use le Chat as your creative partner, teach it to perform specific tasks, use it as a coding copilot… basically, use it to get stuff done.

Today, we’re giving le Chat new powers (in beta). Here’s a quick summary:

Web search with citations

Web search with citations

Canvas for ideation, in-line editing, and export

Canvas for ideation, in-line editing, and export

State of the art document and image understanding, powered by the new multimodal Pixtral Large

State of the art document and image understanding, powered by the new multimodal Pixtral Large

Image generation, powered by Black Forest Labs Flux Pro

Image generation, powered by Black Forest Labs Flux Pro

Fully integrated offering, from models to outputs

Fully integrated offering, from models to outputs

Faster responses powered by speculative editing

Faster responses powered by speculative editing

Here’s how le Chat compares to leading chat and search assistants in the market today:

Feature ChatGPT Perplexity Claude Le Chat Access to latest frontier model Paid Yes Yes Yes Web search with citations Yes Yes No Yes Canvas for ideation, in-line editing, and export Paid No Limited Yes Image generation Yes No No Yes Image / document analysis Yes Paid Yes Yes Task agents Paid No No Yes Single platform from models to outputs Yes No Yes Yes

Feature

ChatGPT

Perplexity

Claude

Le Chat

Access to latest frontier model

Paid

Yes

Yes

Yes

Web search with citations

Yes

Yes

No

Yes

Canvas for ideation, in-line editing, and export

Paid

No

Limited

Yes

Image generation

Yes

No

No

Yes

Image / document analysis

Yes

Paid

Yes

Yes

Task agents

Paid

No

No

Yes

Single platform from models to outputs

Yes

No

Yes

Yes

Let’s dive in.

A significant share of le Chat usage comes from students and professionals who use it for learning, research, and work. To help better facilitate these use cases, we’ve augmented le Chat’s training knowledge with internet search.

Here’s one of our marketing teammates using it to assess the AI market for healthcare.

With state of the art language models that have access to the latest knowledge and happenings, you have a powerful tool to level up your work.

Canvas is a new interface that pops up in your le Chat window when you need to go beyond conversations and into ideation. It enables you to collaborate with Mistral LLMs on shared outputs, enabling 0-1 creation and ideation for knowledge work.

No longer restricted to bidirectional conversations, Canvas enables you to directly modify, edit or transform content aided by the powerful reasoning capabilities of LLMs.

Here’s a video of our product team creating a react mockup of a product metrics dashboard in preparation for the launch of Canvas.

You can use Canvas to create documents, presentations, code, mockups… the list goes on. As you can see, you’re able to modify its contents in place without regenerating responses, version your drafts, and preview your designs.

Canvas allows for infinite creative ideation for any team.

Le Chat is now able to process large, complex PDF documents and images, for analysis and summarization. Simply attach documents to your chat bar and get insights on anything—graphs, tables, diagrams, text, formulae, equations… basically, the entirety of the document.

These advanced capabilities are made possible by the new Pixtral Large, a state of the art multimodal model that is significantly more powerful than our previous vision model, Pixtral 12B , in conjunction with an experimental model (more on that in a few weeks).

In the example below, we show the extraction of information, summarization, and semantic understanding of the famous quantum entanglement paper from 1935 authored by Einstein, Podolsy, and Rosen.

We have partnered with our friends at Black Forest Labs to bring their leading image generation models to le Chat. You can now generate high quality images for your work right from within le Chat.

If you find yourself reusing prompts, or if you’ve identified certain patterns of creative work that you prefer to be automated, you can codify these workflows as agents. Popular use cases include receipt scanning for expense reporting, creating TL;DR’s of long meeting minutes, and invoice processing.

Agents can be published and shared with your teammates. You can ideate recursively, and build increasingly powerful agents that help you get your work done faster. For more on agents, check out this blog or our documentation .

Currently, all of these features are introduced as free beta offerings. At Mistral AI, our approach to AI is different – we’re not chasing AGI at all costs; our mission is to instead place frontier AI in your hands , so you get to decide what to do with advanced AI capabilities. This approach has allowed us to be quite frugal with our capital, while consistently delivering frontier capabilities at affordable price points. On le Chat, we’re offering a generous free tier with these beta features and working towards premium tiers with higher service guarantees.

Head over to chat.mistral.ai to try le Chat. If your current AI-productivity toolchain is siloed or expensive, le Chat will give you an alternative to do your best work. All new features will be rolled out gradually over the next few weeks—be sure to let us know what you think!

0%