Kimi K2: China's New Open-Source AI Model Challenges Global Competitors

'Another DeepSeek moment': Chinese AI model Kimi K2 stirs excitement

Excitement is growing among researchers about a second, powerful artificial intelligence (AI) model to emerge from China, after DeepSeek shocked the world with its launch of R1 in January. The performance of Kimi K2, launched on 11 July by Beijing-based company Moonshot AI, matches or surpasses that of Western rivals, as well as some DeepSeek models, across various benchmarks, according to the firm. In particular it appears to excel at coding, scoring high in tests such as LiveCodeBench, a way of evaluating AIs that challenges models on code-related tasks. As with DeepSeek's models, Kimi K2 is open-weight, meaning it can be downloaded and built upon by researchers for free. It can also be accessed via an application programming interface (API) for a fraction of the price of leading proprietary models, such as Claude 4 from Anthropic, based in San Francisco, California. "The community can freely use it, fine tune it and build on it without training their own model from scratch," says Adina Yakefu, an AI researcher at the open-science platform Hugging Face. Just one day after its launch, Kimi K2 became downloaded at a rate higher than any other model on the platform, Hugging Face data show. Its release is "another 'DeepSeek moment'", Yakefu says. Unlike many other powerful models, K2 is not a 'reasoner' -- a model trained to approach queries using step-by-step logic. Instead, it specialises in being an agentic large language model (LLM), meaning it can carry out multi-step tasks and to do so select from a variety of tools, for example browsing the web or calling upon mathematics software. Proprietary models, such as some versions of ChatGPT, can already do this. But this would be the first time such abilities are available in an open-weight model, says Huan Sun, an AI researcher at Ohio State University in Columbus. However, she cautions that K2 does not appear to be a one-stop-shop for these abilities. "You still need to put it in an agent framework," she says. Having a second impressive model emerge from China within six months suggests that the feat was not an anomaly. "The DeepSeek R1 release earlier this year was more of a prequel than a one-off fluke in the trajectory of AI," wrote Nathan Lambert, a machine learning researcher at the Allen Institute for AI, in his newsletter, Interconnects. Kimi K2 is "the new best open model in the world", he posted on the social media site Bluesky. Moonshot AI, founded in March 2023, is a start-up that until now has been little known in the west. But its Kimi chatbot, based on a previous LLM, was already the third most used in China in November 2024, according to Counterpoint, a marketing research firm based in Hong Kong. Chinese tech giants Alibaba and Tencent are reportedly among its investors. The Kimi K2 model is as hefty as its backers, with one trillion parameters -- the term for the adjustable values that denote the strength of associations within the model. This many parameters would be very challenging for smaller labs to run, says Lambert. However, K2 only activates 32 billion parameters at a time using a 'mixture-of-experts' architecture that allows it to use only the relevant parts of the model for each task, which helps to temper the amount of computing power it requires. As well as coding, Kimi K2 seems to have a flair for writing. Some AI commenters on the social media platform X praised its writing style for sounding unlike a typical AI. The model currently tops the leaderboard on the Creative Writing v3 benchmark, which is designed to test criteria such as the authenticity of characters and avoidance of clichés, and the EQ-bench 3, which examines models' emotional intelligence in role-plays. But K2 does not excel at every task. On SciMuse -- a benchmark that evaluates how well AIs predict which ideas human researchers will find interesting -- it came in behind cutting-edge Gemini algorithms from Google and OpenAI's suite of reasoning models, says Mario Krenn, who leads the Artificial Scientist Lab at the Max Planck Institute for the Science of Light in Erlangen, Germany. Still, Moonshot AI is one of several Chinese firms deciding to publish their models openly, says Yakefu. The United States needs an open model of the calibre of those being produced by DeepSeek and Moonlight AI to counteract the country's diminishing influence in open-source and academic communities, adds Lambert, something he refers to as the American DeepSeek Project. "It's very clear that a large number of top machine learning researchers and engineers, with exceptional hardware, have been behind this effort," says Krenn. "I wouldn't be surprised if more will come [from China] in the next months."

China's Moonshot AI releases open-source model to reclaim market position

BEIJING, July 11 (Reuters) - Chinese artificial intelligence startup Moonshot AI released a new open-source AI model on Friday, joining a wave of similar releases from local rivals, as it seeks to reclaim its position in the competitive domestic market. The model, called Kimi K2, features enhanced coding capabilities and excels at general agent tasks and tool integration, allowing it to break down complex tasks more effectively, the company said in a statement. Moonshot claimed the model outperforms mainstream open-source models in some areas, including DeepSeek's V3, and rival capabilities of leading U.S. models such as those from Anthropic in certain functions such as coding. The release follows a trend among Chinese companies toward open-sourcing AI models, contrasting with many U.S. tech giants like OpenAI and Google that keep their most advanced AI models proprietary. Some American firms, including Meta Platforms (META.O), opens new tab, have also released open-source models. Open-sourcing allows developers to showcase their technological capabilities and expand developer communities as well as their global influence, a strategy likely to help China counter U.S. efforts to limit Beijing's tech progress. Other Chinese companies that have released open-source models include DeepSeek, Alibaba (9988.HK), opens new tab, Tencent (0700.HK), opens new tab and Baidu (9888.HK), opens new tab. Founded in 2023 by Tsinghua University graduate Yang Zhilin, Moonshot is among China's prominent AI startups and is backed by internet giants including Alibaba. The company gained prominence in 2024 when users flocked to its platform for its long-text analysis capabilities and AI search functions. However, its standing has declined this year following DeepSeek's release of low-cost models, including the R1 model launched in January that disrupted the global AI industry. Moonshot's Kimi application ranked third in monthly active users last August but dropped to seventh place by June, according to aicpb.com, a Chinese website that tracks AI products. Reporting by Liam Mo and Brenda Goh, Editing by Louise Heavens Our Standards: The Thomson Reuters Trust Principles., opens new tab Suggested Topics:Media & Telecom

Another Chinese AI model is turning heads

BEIJING -- The latest Chinese generative artificial intelligence model to take on OpenAI's ChatGPT is offering coding capabilities -- at a lower price. Alibaba-backed startup Moonshot released on late Friday night its Kimi K2 model: a low-cost, open source large language model -- the two factors that underpinned China-based DeepSeek's industry disruption in January. Open-source technology provides source code access for free, an approach that few U.S. tech giants have taken, other than Meta and Google to some extent. Coincidentally, OpenAI CEO Sam Altman announced early Saturday that there would be an indefinite delay of its first open-source model yet again due to safety concerns. OpenAI did not immediately respond to a CNBC request for comment on Kimi K2. One of Kimi K2's strengths is in writing computer code for applications, an area in which businesses see potential to reduce or replace staff with generative AI. OpenAI's U.S. rival Anthropic focused on coding with its Claude Opus 4 model released in late May. In its release announcement on social media platforms X and GitHub, Moonshot claimed Kimi K2 surpassed Claude Opus 4 on two benchmarks, and had better overall performance than OpenAI's coding-focused GPT-4.1 model, based on several industry metrics. "No doubt [Kimi K2 is] a globally competitive model, and it's open sourced," Wei Sun, principal analyst in artificial intelligence at Counterpoint, said in an email Monday. "On top of that, it has lower token costs, making it attractive for large-scale or budget-sensitive deployments," she said. The new K2 model is available via Kimi's app and browser interface for free unlike ChatGPT or Claude, which charge monthly subscriptions for their latest AI models. Kimi is also only charging 15 cents for every 1 million input tokens, and $2.50 per 1 million output tokens, according to its website. Tokens are a way of measuring data for AI model processing. In contrast, Claude Opus 4 charges 100 times more for input -- $15 per million tokens -- and 30 times more for output -- $75 per million tokens. Meanwhile, for every one million tokens, GPT-4.1 charges $2 for input and $8 for output. Moonshot AI said on GitHub that developers can use K2 however they wish, with the only requirement that they display "Kimi K2" on the user interface if the commercial product or service has more than 100 million monthly active users, or makes the equivalent of $20 million in monthly revenue. Initial reviews of K2 on both English and Chinese social media have largely been positive, although there are some reports of hallucinations, a prevalent issue in generative AI, in which the models make up information. Still, K2 is "the first model I feel comfortable using in production since Claude 3.5 Sonnet," Pietro Schirano, founder of startup MagicPath that offers AI tools for design, said in a post on X. Moonshot has open sourced some of its prior AI models. The company's chatbot surged in popularity early last year as China's alternative to ChatGPT, which isn't officially available in the country. But similar chatbots from ByteDance and Tencent have since crowded the market, while tech giant Baidu has revamped its core search engine with AI tools. Kimi's latest AI release comes as investors eye Chinese alternatives to U.S. tech in the global AI competition. Still, despite the excitement about DeepSeek, the privately-held company has yet to announce a major upgrade to its R1 and V3 model. Meanwhile, Manus AI, a Chinese startup that emerged earlier this year as another DeepSeek-type upstart, has relocated its headquarters to Singapore. Over in the U.S., OpenAI also has yet to reveal GPT-5. Work on GPT-5 may be taking up engineering resources, preventing OpenAI from progressing on its open-source model, Counterpoint's Sun said, adding that it's challenging to release a powerful open-source model without undermining the competitive advantage of a proprietary model. Kimi K2 is not the company's only recent release. Moonshot launched a Kimi research model last month and claimed it matched Google's Gemini Deep Research 's 26.9 score and beat OpenAI's version on a benchmark called "Humanity's Last Exam." The Kimi research model even got a mention last week during Elon Musk's xAI release of Grok 4 -- which scored 25.4 on its own on the "Humanity's Last Exam" benchmark, but attained a 44.4 score when allowed to use a variety of AI tools and web search. "Kimi-Researcher represents a paradigm shift in agentic AI," said Winston Ma, adjunct professor at NYU School of Law. He was referring to AI's capability of simultaneously making several decisions on its own to complete a complex task. "Instead of merely generating fluent responses, it demonstrates autonomous reasoning at an expert level -- the kind of complex cognitive work previously missing from LLMs," Ma said. He is also author of "The Digital War: How China's Tech Power Shapes the Future of AI, Blockchain and Cyberspace."

China's Moonshot launches open source AI model, OpenAI postpones

Moonshot is the latest Chinese start-up to launch a new open-source AI model onto the market in the shape of Kimi 2, after DeepSeek shook the market back in January. Alibaba-backed Moonshot AI hopes to take on Deepseek and other local Chinese rivals - as well as it's US counterparts - with its the open-source large language model, Kimi 2. The launch coincides with OpenAI's Sam Altman announcing this weekend that its first open-source model would be delayed indefinitely for safety reasons. Moonshot says that in testing it beat ChatGPT and Anthopic's Claude in coding, and at a lower cost. Kimi-K2-Base is the foundation model, "for researchers and builders who want full control for fine-tuning and custom solutions", while Kimi-K2-Instruct is the "post-trained model best for drop-in, general-purpose chat and agentic experiences", according to Moonshot AI. Moonshot AI says the model outperforms mainstream open-source models in certain areas, including DeepSeek's V3 and Anthropic's Claude in areas like coding. While Meta Platforms has released an open source model, the majority of US players keep their large AI models proprietary, so many see the open source model being practiced by many Chinese companies as a way to engage more deeply with the developer community, and compete with their US rivals. Tsinghua University graduate Yang Zhilin founded Moonshot in 2023 and it has influential backers including Chinese e-commerce giant Alibaba. It had strong initial success with Kimi, thanks in part to its AI search functions and long text analysis, but Deepseek's release in January saw it lose ground. Now, it's hoping to grab the limelight again with the latest release, which outperforms ChatGPT and Claude in certain functions on the industry benchmarks like SWE-bench Verified. The real winner for Kimi 2 will likely be its so called 'agentic' abilities - the ability to accomplish multi-task goals with limited human intervention. Kimi 2 makes Moonshot AI the latest player to seriously up the competition among Chinese and US AI models. Don't miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic's digest of need-to-know sci-tech news.

China's Kimi K2 Could Be the Next DeepSeek Moment | AIM

As soon as the model was dropped, OpenAI announced a delay in the release of its open-source model. China's open-source AI scene is heating up again. After DeepSeek's rapid rise earlier this year, a new challenger is making waves in the form of Kimi K2 from Moonshot AI. Although it launches with less fanfare, Kimi K2 is now drawing serious attention from AI insiders and outperforming some of the biggest names in the game. It's fast, climbing the ranks, beating expectations on benchmarks, and sparking comparisons to DeepSeek's breakout moment. Some even believe it's strong enough to have made OpenAI rethink its release schedule. "China's Kimi K2 is having its mini DeepSeek moment: it is now #14 on OpenRouter today, ahead of Grok 4 and GPT-4.1," Deedy Das of Menlo Ventures wrote in a post on X He added that this is a non-reasoning model, yet it scores highest on major EQ and creative writing benchmarks. "Best model smell since (Claude) 3.5 Sonnet," he said. Based on current API pricing, Kimi K2 is roughly 80-90% cheaper than Claude Sonnet 4 when comparing per-token costs, especially for API usage. Kimi K2 uses a sparse mixture‑of‑experts (MoE) design, featuring one trillion total parameters and 32 billion active ones per query. Of its 384 specialised expert subnetworks, only a few are activated dynamically based on the input. This setup lowers compute needs while preserving capacity. It also supports a 1,28,000-token context window. As soon as the model was dropped, OpenAI CEO Sam Altman announced a delay in the release of their open-source model. "Kimi mogged OpenAI, and I genuinely think the real reason they delayed the open-source model release is Kimi K2," AI enthusiast Ashutosh Shrivastava wrote on X. He added that OpenAI "never saw this coming". Kimi K2 outperforms DeepSeek V3 and goes head-to-head with Claude Opus 4 and GPT-4.1. This comes against the backdrop of OpenAI naming another Chinese AI startup, Zhipu, as a potential threat to its dominance. Kimi K2 delivered top-tier results in coding and math benchmarks. On SWE-bench Verified, it scored 65.8%, outperforming GPT-4.1 at 54.6% and coming close to Claude Sonnet 4. On LiveCodeBench, it achieved 53.7%, ahead of DeepSeek V3 (46.9%) and GPT-4.1 (44.7%). In the Math-500 benchmark, it scored 97.4%, compared to GPT-4.1's 92.4%. Kimi K2 also performs strongly across AIME, GPQA, OGBench, and tool-use evaluations. Artificial Analysis said that while Moonshot AI's Kimi K2 is the leading open-weight non-reasoning model in its Intelligence Index, it outputs roughly three times more tokens than other non-reasoning models, blurring the line between reasoning and non-reasoning. As a non-reasoning model, it excels in creative tasks. It is now the Short-Story Creative Writing champion, scoring 8.56 and surpassing the previous leader, o3-pro, which scored 8.44. Kimi K2 has good agentic capabilities. According to the company, unlike traditional LLMs, Kimi K2 can plan and execute multi‑step tasks autonomously. It can call external APIs, generate and debug code, create plots, webpages and more, all without manual prompting at each step. There are two versions of the model. While the Base variant is designed for research and fine-tuning, the Instruct variant is intended for use in chatbots and agents. In a blog post, the company shared that Kimi K2's agentic abilities are driven by two core components: large-scale tool-use training and general reinforcement learning (RL). In order to teach the model how to use tools effectively, Moonshot AI built a large-scale synthetic data pipeline inspired by ACEBench. This system simulates real-world tool-use tasks across hundreds of domains and thousands of tools, combining both real and synthetic examples. "Our approach systematically evolves hundreds of domains containing thousands of tools, including both real MCP (Model Context Protocol) tools and synthetic ones, then generates hundreds of agents with diverse tool sets," the company said. Despite the good benchmark figures, Ethan Mollick, a professor at Wharton, described Kimi K2 as "a really weird model" that still needs much more testing. He recounted an experiment where he gave it a slightly altered version of the novel The Great Gatsby. Like Claude, the model spotted the two intentional changes, but then "made up a ton of hallucinated nonsense that sounded plausible but was just plain wrong". He added that the DeepSeek moment was largely fueled by pent-up consumer demand for high-quality free AI, especially among students looking for help with homework. According to him, Kimi K2, despite its strong performance, hasn't seen the same immediate public impact. One possible reason he observed is that for most consumers and students, "DeepSeek is good enough". "Feels like unlike DeepSeek, the general public hasn't felt the effect/impacts of Kimi K2 yet - most non-technical people have probably never even heard of it. Wonder why it is being overlooked when DeepSeek got so much attention," wrote a user on X. Meanwhile, DeepSeek's upcoming model, R2, is still unreleased, and it may be delayed further. A recent report suggests that US export restrictions on NVIDIA's H20 chips, which are essential for training and deploying the model, could pose serious challenges in China. Kimi K2 may not have the same hype DeepSeek had, but its performance is hard to ignore. With strong benchmarks and growing visibility, it is clear that China's open-source push is far from over.

China's Moonshot AI releases open-source model to reclaim market position - The Economic Times

The model, called Kimi K2, features enhanced coding capabilities and excels at general agent tasks and tool integration, allowing it to break down complex tasks more effectively, the company said in a statement.Chinese artificial intelligence startup Moonshot AI released a new open-source AI model on Friday, joining a wave of similar releases from local rivals, as it seeks to reclaim its position in the competitive domestic market. The model, called Kimi K2, features enhanced coding capabilities and excels at general agent tasks and tool integration, allowing it to break down complex tasks more effectively, the company said in a statement. Moonshot claimed the model outperforms mainstream open-source models in some areas, including DeepSeek's V3, and rival capabilities of leading U.S. models such as those from Anthropic in certain functions such as coding. The release follows a trend among Chinese companies toward open-sourcing AI models, contrasting with many U.S. tech giants like OpenAI and Google that keep their most advanced AI models proprietary. Some American firms, including Meta Platforms, have also released open-source models. Open-sourcing allows developers to showcase their technological capabilities and expand developer communities as well as their global influence, a strategy likely to help China counter U.S. efforts to limit Beijing's tech progress. Other Chinese companies that have released open-source models include DeepSeek, Alibaba, Tencent and Baidu. Founded in 2023 by Tsinghua University graduate Yang Zhilin, Moonshot is among China's prominent AI startups and is backed by internet giants including Alibaba. The company gained prominence in 2024 when users flocked to its platform for its long-text analysis capabilities and AI search functions. However, its standing has declined this year following DeepSeek's release of low-cost models, including the R1 model launched in January that disrupted the global AI industry. Moonshot's Kimi application ranked third in monthly active users last August but dropped to seventh place by June, according to aicpb.com, a Chinese website that tracks AI products.

Kimi K2 One Trillion Parameter Open Source AI Model Tested

What if the future of artificial intelligence wasn't locked behind corporate walls but instead placed directly in the hands of developers, researchers, and innovators worldwide? Enter Kimi K2, a new open source AI model from Moonshot that's shaking up the status quo. With its one trillion parameter mixture of experts architecture, Kimi K2 doesn't just rival proprietary systems -- it redefines what's possible for accessible AI. Imagine a model that can autonomously reason, execute multi-step tasks, and seamlessly integrate external tools, all while maintaining computational efficiency. This isn't just another AI release; it's a bold statement about the future of provide widespread access tod technology. In this exploration, World of AI uncover how Kimi K2's innovative design and state-of-the-art performance are setting new benchmarks in the AI landscape. From its ability to outperform competitors in STEM challenges to its versatility across industries like healthcare, education, and gaming, Kimi K2 is more than just a model -- it's a movement. Whether you're a developer seeking fine-tuned control or a user looking for an out-of-the-box solution, Kimi K2's dual versions promise something for everyone. But what does this mean for the dominance of closed-source AI systems? Let's unpack the implications of this innovative release and what it signals for the future of artificial intelligence. Kimi K2's architecture is built on a mixture of experts system, a design that activates only the most relevant parameters for each task. This selective activation ensures computational efficiency while maintaining high performance. The model is specifically optimized for agentic reasoning, allowing it to autonomously make decisions and execute multi-step workflows. Its ability to integrate external tools further enhances its adaptability, making it suitable for addressing complex, resource-intensive tasks across various domains. This architecture not only reduces computational overhead but also allows Kimi K2 to scale effectively for diverse applications. By focusing on task-specific parameter activation, the model achieves a balance between efficiency and accuracy, setting it apart from traditional AI systems. Kimi K2 delivers state-of-the-art results across several industry-standard benchmarks, including Swaybench, GSM8K, and AceBench Math. These evaluations underscore its strengths in reasoning, problem-solving, and computational tasks. For example: These results position Kimi K2 as a credible and capable alternative to closed-source AI systems, offering comparable or superior performance in key areas. To meet the diverse requirements of its users, Kimi K2 is available in two distinct versions, each designed for specific use cases: These options ensure that both advanced developers and general users can use Kimi K2's capabilities effectively, whether for research, development, or practical applications. Kimi K2's adaptability makes it suitable for a wide range of applications across industries. For web developers, it can generate high-quality front-end designs and functional layouts, streamlining the design process. Its ability to create SVG representations and analyze visual data provides actionable insights through graphical outputs, enhancing decision-making and design workflows. In more advanced scenarios, Kimi K2 has been employed to develop intricate 3D simulations, such as a Minecraft-like environment. This demonstrates its capacity to handle complex and creative tasks, making it a valuable tool for developers exploring virtual environments, gaming, and simulation-based applications. Beyond development, Kimi K2's capabilities extend to fields such as education, where it can assist in creating interactive learning tools, and healthcare, where it can analyze large datasets to identify trends and insights. Its versatility ensures that it can adapt to the unique demands of various industries. One of Kimi K2's most compelling features is its accessibility. As an open source model, users can download and deploy its weights locally, granting them full control over its implementation. This approach enables developers and organizations to customize the model to suit their specific needs without relying on external infrastructure. For those who prefer cloud-based solutions, Kimi K2 offers an API with competitive pricing for input and output tokens. To further reduce barriers to entry, Moonshot provides free API credits for initial usage, allowing users to explore the model's potential without incurring upfront costs. This combination of flexibility and affordability makes Kimi K2 an attractive option for businesses and individuals alike. Kimi K2 represents a significant milestone in the evolution of artificial intelligence. By offering a high-performance, open source alternative to proprietary models, it provide widespread access tos access to innovative AI technologies. Its combination of cost-effectiveness, versatility, and robust performance positions it as a valuable resource for businesses, researchers, and developers. Whether your goal is to optimize workflows, create innovative applications, or explore the frontiers of AI, Kimi K2 provides a powerful and accessible solution. Its release not only challenges the status quo but also paves the way for a more inclusive and collaborative AI ecosystem, where advanced technology is within reach for all.

China's Moonshot AI releases open-source model to reclaim market position

BEIJING (Reuters) -Chinese artificial intelligence startup Moonshot AI released a new open-source AI model on Friday, joining a wave of similar releases from local rivals, as it seeks to reclaim its position in the competitive domestic market. The model, called Kimi K2, features enhanced coding capabilities and excels at general agent tasks and tool integration, allowing it to break down complex tasks more effectively, the company said in a statement. Moonshot claimed the model outperforms mainstream open-source models in some areas, including DeepSeek's V3, and rival capabilities of leading U.S. models such as those from Anthropic in certain functions such as coding. The release follows a trend among Chinese companies toward open-sourcing AI models, contrasting with many U.S. tech giants like OpenAI and Google that keep their most advanced AI models proprietary. Some American firms, including Meta Platforms, have also released open-source models. Open-sourcing allows developers to showcase their technological capabilities and expand developer communities as well as their global influence, a strategy likely to help China counter U.S. efforts to limit Beijing's tech progress. Other Chinese companies that have released open-source models include DeepSeek, Alibaba, Tencent and Baidu. Founded in 2023 by Tsinghua University graduate Yang Zhilin, Moonshot is among China's prominent AI startups and is backed by internet giants including Alibaba. The company gained prominence in 2024 when users flocked to its platform for its long-text analysis capabilities and AI search functions. However, its standing has declined this year following DeepSeek's release of low-cost models, including the R1 model launched in January that disrupted the global AI industry. Moonshot's Kimi application ranked third in monthly active users last August but dropped to seventh place by June, according to aicpb.com, a Chinese website that tracks AI products. (Reporting by Liam Mo and Brenda Goh, Editing by Louise Heavens)