Moonshot Kimi K3
Moonshot's 2.8T-parameter open-weight mixture-of-experts reasoning model (104B active) with a 1M-token context, built for long-horizon agent work, coding and research.
From Moonshot, Kimi K3 is a cloud text model. It is multimodal: alongside a text prompt it accepts images, then replies with generated text. Its context window handles up to 1.05M input tokens and up to 131K output tokens, and it supports adjustable reasoning effort for harder problems. Generation can be tuned with system instructions, temperature, and top-p sampling. It runs through Runware and OpenRouter using your own API key, from $0.239 per million input tokens.
- Released
- Jul 2026
- Pricing
- $0.239 / 1M in · $10.80 / 1M out
- Inputs
- Text, Images
- Context window
- 1.05M in · 131K out
- Reasoning
- Adjustable effort
- Controls
- System prompt, Temperature, Top-p
Samples
Generated with Kimi K3 via Runware. The same three prompts run against every text model in the catalog, shown verbatim, so the only thing that changes between two models’ answers is the model.
Explain what an API is to someone who has never written code. Do not use a restaurant or waiter analogy. Plain prose only — no headings, bullet points, or markdown formatting. Maximum 50 words.
An API is a set of rules that lets one computer program ask another for something. Like a light switch — you flip it and the light comes on without understanding the wiring — a program can request weather data or process a payment expecting a predictable result, no inside knowledge needed.
A two-person startup is deciding whether to build their own user authentication or use a hosted provider. Give them a clear recommendation, then the single strongest argument against your own recommendation. Plain prose only — no headings, bullet points, or markdown formatting. Maximum 150 words.
Use a hosted provider. With two people, your scarcest resource is engineering time, and authentication is a solved problem that is dangerous to get subtly wrong: password storage, session handling, token refresh, MFA, and account recovery all have to be correct simultaneously, and mistakes only surface when you get breached or lock users out. Providers like Clerk, Auth0, or Supabase Auth give you all of this for a modest fee, letting you ship your actual product faster, which is the only thing that matters at your stage. The strongest argument against: lock-in. Your users' credentials live on someone else's infrastructure, priced per monthly active user, so success makes the bill grow and leaving becomes painful. Migrating off a provider often means forcing every user through a password reset, and any outage or abrupt pricing change on their end instantly becomes your outage and your problem, with little recourse.
Write a briefing for a non-technical executive explaining why their company's spending on AI is hard to predict, and what they can do about it. Cover at least three distinct causes. Plain prose only — no headings, bullet points, or markdown formatting. Write exactly 205 words.
There are several reasons your company's AI spending is difficult to forecast, and understanding them will help you budget with greater confidence. First, most AI services bill like utilities, charging per unit of work processed rather than a flat monthly license. Costs therefore swing with how much employees actually use the tools, which is nearly impossible to predict in advance. Second, the market itself is unstable. Vendors change prices frequently, retire older models, and release new ones with very different cost structures, so a budget built on January's assumptions may be obsolete by summer. Third, the invoice is only part of the expense. Preparing data, connecting tools to existing systems, checking outputs for errors, and reworking failures often cost more than the technology itself, and these expenses surface late. Fourth, success drives spending: when a tool works well, adoption spreads and bills climb. You can manage this uncertainty. Set spending caps and usage alerts so surprises surface quickly. Fund small pilots with clear success measures before approving large rollouts. Assign one senior owner to review total AI spending monthly, and negotiate committed-use discounts once usage patterns stabilize. With these guardrails, AI becomes a manageable operating expense rather than an unpredictable item undermining your annual planning.
Moonshot
Moonshot builds the Kimi family — frontier open-weight models with very long context, native visual understanding, and strong agentic coding.
www.moonshot.ai ↗Moonshot's flagship open-weight multimodal reasoning model, post-trained from Kimi K3 for long-horizon coding and agent work, with a 1M-token context.
Moonshot's frontier open-weight model (1T parameters) built for long-horizon agentic coding, with a 262K context and native vision.