Kimi

On November 9, Moonshot AI launched its most powerful open-source thinking model so far on Thursday. According to the dark side of the month, Kimi K2 Thinking achieved an excellent performance of 44.9% in the final human examination (HLE), which went beyond advanced models such as GPT-5, Grok-4, Claude 4.5. However, CNBC quotes informed sources that Kimi K2 Tinki..

On November 7th, the dark side of the moon launched Kimi's most powerful open-source thinking model - Kimi K2 Thinking. The model was described as a new generation of Tinking Agent, trained in the concept of the "model is Agent" on the dark side of the moon, where originals mastered the ability to "think and use tools". In the final human examination (Humanity & #039; s Last Exam), autonomous web browse Comp, complex information gathering reasoning (SEAL-0)

On September 26, Kimi, the dark side of the moon, released a new Agent mode OK Computer, starting the greyscale test. OK Computer continues the concept of "model is Agent" and further enhances the ability of smarts and tools to access through end-to-end training of Kimi K2. After the user's demand, Kimi can operate his own virtual computer and perform complex tasks such as multifunctional website development, big data analysis, photo video generation and high quality PPT production. According to 1Ai, there was a reward..

On August 23rd, Dark Side of the Moon announced that through the tireless efforts of its engineers, the output speed of the kimi-k2-turbo-preview model has been increased to 60 Tokens per second, with a maximum of 100 Tokens per second. Currently, the model is still enjoying a 50% discounted price, with the model costing ¥2.00 per million tokens for the input price (cache hits), ¥8.00 for the input price (cache misses) and ¥32.00 for the output price. Input price (cache hit) ¥2.00, input price (cache miss) ¥8.00, output price ¥32.00. The original price will be restored on September 1st. As 1AI previously reported, Dark Side of the Moon 8 ...

August 1, news, the dark side of the moon today released Kimi K2 high-speed version -- Kimi-K2-turbo-preview, the model parameters and Kimi-K2 consistent, but the output speed from 10 Tokens per second to 40 Tokens per second. According to the introduction of the model is currently a limited time 50% off! According to the report, the model is currently on special offer for a limited time at 50% off, and the original price will be restored on September 1. The discounted price: the model is priced at $2 per million tokens for input (cache hits), $8 for input (cache misses), and $32 for output. According to 1...

July 19, 2011 - LMArena, the leading ranking of big models, has released its latest rankings, with the recently released Kimi K2 taking the top spot over DeepSeek R1 for open source models. According to LMArena, Kimi K2 took fifth place in LMArena's overall ranking based on its performance and 3,000 community votes. It's worth noting that Kimi K2 and DeepSeek R1 are the two Chinese models in the top 10 of the LMArena charts, but...

July 12, 2012 - Dark Side of the Moon officially released the Kimi K2 model last night and open-sourced it simultaneously. Kimi K2 is a MoE architecture base model with stronger code capabilities and better at generalized Agent tasks, with 1T total parameters and 32B activation parameters. In benchmark performance tests such as SWE Bench Verified, Tau2, AceBench, etc., Kimi K2 has achieved SOTA scores among the open-source models, demonstrating its leading capabilities in code, Agent, mathematical reasoning tasks. ...

Have you ever done the math: how much precious energy is spent on making a PPT? Often friends complain to me that PPT is troublesome to do, most of the time is wasted on adjusting the format, good-looking templates have to spend money to buy, and the free ones are too monotonous. ...... There is a saying that the hardest thing about PPT is not the thinking, but the expression; the annoying thing is not the content, but the package. Looking for templates, aligning icons, matching colors, there is this time, whether it is to write one more manuscript, or to enjoy life a little more, it is much more practical. Don't worry, AI This is not coming, if you...

June 21, 1AI learned from Kimi's public number on the dark side of the moon that Kimi-Researcher, Kimi's first Agent (intelligent body), opened a small-scale gray-scale test on the 20th. Kimi-Researcher is a new generation of Agent model trained based on end-to-end autonomous reinforcement learning (end-to-end agentic RL) technology, and is also an Agent product designed for deep research tasks. Later, Dark Side of the Moon will gradually open source Kimi-Res...

Still struggling with which AI tool to choose? DeepSeek, Gemini, Tongyi Thousand Questions, Beanbag, Kimi... Mainstream AI tools on the market are blossoming, with different functions and dazzling publicity and selling points. But there is no "best", only "most suitable". What kind of AI partner do you need for your creative inspiration, office efficiency, code development or knowledge exploration? In this article, we clearly explain the unique advantages and potential shortcomings of the five major tools from the core capabilities, specific scene performance, interactive experience to integration and scalability and other key dimensions. After reading this article, you may no longer be confused and pinpoint...

Today on a full range, hand in hand to teach you to use Kimi to complete the creation of the screenplay script, as well as the subsequent text of the picture, picture video, and even background music creation all unlocked, to achieve the full process of production of the Wong Kar-wai style movie. From the script to the finished film in just 6 steps: The first step, Kimi generates split-screen scripts Kimi website: https://www.1ai.net/3878.html First of all, to teach Kimi, so that it generates split-screen scripts more in line with our requirements. First, let it understand the style of Wong Kar-wai's movie "I want to reproduce the style of director Wong Kar-wai's movie to produce...

April 7, Kimi open platform today announced that based on Moonshot AI's technology accumulation and performance optimization over the past year, it has adjusted the price of model inference services provided by Kimi open platform at 0:00 on April 07, 2025, Beijing time, the specific adjustment plan is as follows: Note: the price of kimi-latest model after automatic caching is still ¥1 / M Kimi Open Platform said: "Through technological innovation, we have successfully reduced the cost of model inference and improved the inference...

Kimi Intelligent Assistant's image creation function allows users to quickly generate images that meet their needs by entering keywords or descriptions. The following are the main features of this function: 1. Generation method Users only need to provide clear keywords or descriptive text, and Kimi will search and integrate relevant image materials according to these prompts to generate images that meet the needs. 2. Support multiple ratios The feature supports multiple image ratios, including 1:1, 3:4, 4:3, 9:16, 16:9, etc., to meet the needs of different scenarios. 3. Editing and Sharing The generated images support further editing...

February 27, 2011 - Kimi's latest model, k1.6, was revealed today. According to LiveCodeBench, a global dynamic benchmarking platform, Kimi k1.6 surpassed GPT o3mini, o1 and other models to reach the top of the code generation test. Flood, a researcher at Dark Side of the Moon, said on social media platforms that k1.6 is based on the training done by k1.5, and the performance is still being improved. Zhang Yutong, co-founder of Dark Side of the Moon, said in his circle of friends, "k1.6 has not been released yet, but seeing LiveCodeBench...

February 24, 2011 - Kimi, Dark Side of the Moon, yesterday released a new technical paper "Muon Scalable for LLM Training" and announced the launch of "Moonlight": a 3 billion / 16 billion parameter hybrid expert model (MoE) trained on Muon. Mixed Expert Model (MoE) trained on Muon. Using 5.7 trillion tokens, it achieves better performance at lower floating point operations (FLOPs), thus improving the Pareto efficiency bound. Dark Side of the Moon says the team found that the Muon optimizer can be carefully tuned by adding weight decay,...

February 18, 2011 - Kimi, a subsidiary of Dark Side of the Moon, was one of the most talked about AI products from startups in China before the rise of DeepSeek. However, after DeepSeek's open source, low-cost, high-performance explosion, some people can't help but compare it to Kimi. According to a report by Interface News today, Dark Side of the Moon has recently decided to drastically reduce its product placement budget, including the suspension of multiple Android channel placements and cooperation with third-party advertising platforms. The report mentioned that in the face of DeepSeek's natural traffic in a few dozen days went to 35 million...

February 18th, 2011 - Dark Side of the Moon has announced the launch of a new model, kimi-latest. kimi-latest will be benchmarked against the current model used by Kimi's Intelligent Assistant, and upgraded in tandem with updates to the Intelligent Assistant product (the model name will always be kimi-latest). Kimi Latest Model Features Kimi Latest always uses the latest Kimi Big Model used by Kimi Intelligent Assistant products, and may contain features that are not yet stable Context length is 128k, and will be automatically upgraded according to the...

February 18th morning news, the dark side of the moon researcher Flood Sung recently shared the complete thinking process behind the k1.5 model, and revealed that the shock effect brought by the release of OpenAI o1 on September 12th, 2024 has plunged himself into the reflection on the effectiveness of Long-CoT. Because the effectiveness of Long-CoT, in fact, more than a year ago, the dark side of the moon Kimi co-founder Tim Zhou Xinyu verified that the use of very small models, training models to do do dozens of digits of addition, subtraction, multiplication and division, will be fine-grained...

As you may have seen recently, DeepSeek, the world's most popular AI tool, added 100 million new users in just 7 days, making it a king-size product. What kind of chemical reaction will be created if we combine this AI tool with another AI tool? Today, let's take a look at how to use DeepSeek and Kimi to generate explosive PPTs with one click! First, open the official website of DeepSeek, https://www.1ai.net/24339.html Ask it a question on the official webpage...

Dark Side of the Moon announced on January 20 the launch of Kimi's new SOTA model -- k1.5 multimodal thinking model, which realizes SOTA (state-of-the-art) level multimodal reasoning and generalized reasoning capabilities. Officially, in short-CoT mode, Kimi k1.5's mathematical, coding, and visual multimodal and generalized capabilities dramatically outperform those of the global short-thinking SOTA models GPT-4o and Claude 3.5 Sonnet, leading by as much as 5...

January 15, 2011 - The Dark Side of the Moon today released the Kimi Multimodal Image Understanding Model API, a new multimodal image understanding model, moonshot-v1-vision-preview (hereinafter referred to as the "Vision model"), which completes the moonshot-v1 model series' multimodal capabilities. model series. Model Capability Description Image Recognition The Vision model has the ability to recognize complex details and nuances in an image, whether it is a food or an animal, and to distinguish between objects that are similar but not identical. ...

A lot of people say: review 2024 annual report, summarize the two words "crazy", not crazy can not, counting a few more days poor, this year even poor to the end, obviously no money to go to work, but found that after the work of the poor more stable, January 30th, every day for the money to turn; a year in December, the pocket is empty, remember the beginning of the year set up the I'm not sure if you're going to be able to get a good deal on a new one, but I'm sure you'll be able to get a good deal on a new one, so I'm sure you'll be able to get a good deal on a new one. Work is not finished, sleep is not enough, money can not earn, weight is soaring, this may be the annual epitome of many people, then why not try to use AI to improve the efficiency of the writing, both to complete some of the work ...

Every morning after sitting in the office seat, open the computer, facing a pile of tasks to be done, do you often do not know where to start? You are not inspired to write copy, you don't have ideas to make PPT, you need to organize meeting minutes.... Sometimes you want AI to help you, but you feel it can't understand your needs? In fact, AI is like a hidden assistant, the key is to use the right 'code'to activate its potential. I summarized the 20+ practical instructions to make AI turn into a super assistant in a second, to help you achieve a qualitative leap in the two dimensions of content creation and workplace efficiency. The content of the instructions is divided into two major...

Do you often struggle to come up with a good line? Are you envious of those big Vs who can come up with a golden sentence at the drop of a hat? Do you want to make your articles full of sparkles, but always feel that you are just a little bit short of the fire? If any of the above bothers you, then congratulations, you've come to the right place! As a blogger who is deeply involved in AI writing, I know the importance of a golden sentence in a great post. A well-crafted golden sentence can often instantly elevate a mediocre article and make it unforgettable for readers. However, creating golden sentences that are both insightful and catchy is not easy. I used to be like...