Moonshot’s Kimi K2 surpasses GPT, Claude in coding benchmark
BEIJING, CHINA — Moonshot, an AI startup backed by Alibaba, has unveiled its latest large language model, Kimi K2, which is revolutionizing the generative AI field by offering high-performance coding capabilities at a fraction of the cost of its competitors.
Kimi K2 is both open-source and low-cost, giving developers free access to its underlying code, an approach seldom seen among major Western AI firms outside of Meta and Google.
The model, available for free via app and browser, is aimed squarely at rival heavyweights like OpenAI’s GPT-4.1 and Anthropic’s Claude Opus 4.
In initial tests claimed by Moonshot, Kimi K2 outperformed Claude Opus 4 on two key coding benchmarks and demonstrated superior overall results to GPT-4.1, according to standard industry metrics.
🚀 Hello, Kimi K2! Open-Source Agentic Model!
🔹 1T total / 32B active MoE model
🔹 SOTA on SWE Bench Verified, Tau2 & AceBench among open models
🔹Strong in coding and agentic tasks
🐤 Multimodal & thought-mode not supported for nowWith Kimi K2, advanced agentic intelligence… pic.twitter.com/PlRQNrg9JL
— Kimi.ai (@Kimi_Moonshot) July 11, 2025
Surpassing rivals in performance and price
One standout feature of Kimi K2 is its ability to generate computer code, addressing business demands for automated coding and productivity.
“No doubt [Kimi K2 is] a globally competitive model, and it’s open sourced,” noted Wei Sun, principal AI analyst at Counterpoint.
Sun added that Kimi K2 “has lower token costs, making it attractive for large-scale or budget-sensitive deployments”.
Kimi K2’s pricing strategy includes:
- Input tokens: $0.15 per million (versus $15 for Claude Opus 4 and $2 for GPT-4.1)
- Output tokens: $2.50 per million (compared to $75 from Claude and $8 from GPT-4.1)
Unlike ChatGPT or Claude, which require subscription fees for their latest models, Kimi K2 can be used for free, with ultra-low charges for commercial developers.
The only stipulation is that apps or services with over 100 million monthly users or $20 million in revenue must visibly credit “Kimi K2.”
AI race intensifies as U.S. firms hit pause
Moonshot’s announcement closely followed news from OpenAI CEO Sam Altman, who on Saturday revealed an indefinite delay for OpenAI’s first open-source model, again citing safety concerns.
we planned to launch our open-weight model next week.
we are delaying it; we need time to run additional safety tests and review high-risk areas. we are not yet sure how long it will take us.
while we trust the community will build great things with this model, once weights are…
— Sam Altman (@sama) July 12, 2025
Counterpoint’s Sun suggested that engineering resources may be shifting to OpenAI’s GPT-5, explaining the company’s slower rollout of its open-source products.
Initial reviews of Kimi K2 on both English and Chinese social media have been mostly positive, though some reports cite occasional “hallucinations”—AI-generated inaccuracies common in the field.
Moonshot is also pushing agentic AI research: last month, its Kimi Researcher model tied Google’s Gemini on a reasoning benchmark known as “Humanity’s Last Exam,” and outperformed OpenAI’s equivalent.
“Kimi-Researcher represents a paradigm shift in agentic AI,” said Winston Ma, adjunct law professor at NYU.
“Instead of merely generating fluent responses, it demonstrates autonomous reasoning at an expert level—the kind of complex cognitive work previously missing from LLMs.”
China’s AI push challenging U.S. dominance
Moonshot’s aggressive move fills a growing demand for affordable, open AI, especially as U.S. players face slowdowns over safety and regulatory issues. As one industry observer summed up: “It’s the first model I feel comfortable using in production since Claude 3.5 Sonnet.”
If this momentum leads to global adoption, it could reshape the AI landscape, while Chinese and American competitors scramble for the lead.
