Kimi API Pricing 2026: K2, K2.6 & Thinking Tiers Compared

Base Kimi K2 (the K2-0711 generation) costs $0.57 in / $2.30 out per million tokens measured via OpenRouter, near parity with Moonshot's official legacy K2 at $0.55 / $2.20, while the current flagship K2.6 is a separate, pricier tier at $0.95 / $4.00 with a 256K context. That one sentence hides the mistake most buyers make: they read a single "Kimi" price and budget against the wrong model. This hub fixes that by naming every tier, pricing each one against the route you actually use, and pointing you to the nine focused articles in this cluster.

Which Kimi are you actually pricing?

"Kimi" is not one model. Moonshot ships a family, and the cheap headline number you see quoted online is usually the base K2 generation, not the flagship people assume they are buying.

Kimi K2 is a large language model family from Moonshot AI that pairs a long context window with separate reasoning and coding variants at different price points. The practical consequence is that the word "Kimi" in a pricing thread is ambiguous until someone names the tier. The OpenRouter moonshotai/kimi-k2 alias resolves to the K2-0711 generation, which is the legacy base tier, not the newer K2.6 flagship. If you price K2.6's capability and pay K2-0711's bill, you have mis-scoped the project.

According to Moonshot AI's platform pricing, the current flagship Kimi K2.6 runs $0.95 input / $4.00 output per million tokens with a 262,144-token (256K) context and a $0.19 cached-input rate. That is the tier to budget against when you need the long-context flagship, and it is materially more expensive than the base K2 alias most cheap-Kimi articles quote.

Kimi tier pricing table (verified 2026-06)

Here is every tier we priced, with the route disclosed for each row. OpenRouter rows are live catalog numbers we read on 2026-06-26; the official Moonshot rows are sourced research pending native re-verification.

A note we are deliberately not glossing over: for base K2, the OpenRouter price ($0.57 / $2.30) sits at a slight premium over the official legacy K2 ($0.55 / $2.20), not a discount. That is the opposite of the pattern some Chinese-model routes show, and we would rather you know it than discover it on an invoice.

According to Kimi/Moonshot pricing, the official legacy K2 lists at $0.55 input and $2.20 output; the OpenRouter-routed base kimi-k2 alias (measured via OpenRouter) sits about two cents per million higher, rounding noise at most workload sizes. The reason to pick a route is therefore access and reliability, not a meaningful price gap on base K2.

First-hand: the K2-thinking reasoning tax (measured via OpenRouter)

Documentation tells you a reasoning tier costs more per token. It does not tell you how much more the tier spends on its own thinking. So we called the endpoints and measured it.

When we sent the same two-sentence general prompt to base K2 and to K2-thinking through OpenRouter on 2026-06-26, base K2 answered with 60 output tokens in 7.03 seconds for $0.00015852. The K2-thinking tier (served as moonshotai/kimi-k2-thinking-20251106) answered the same prompt with 553 output tokens in 18.51 seconds for $0.0014041. Same question, roughly 9x the cost and more than double the latency, almost entirely because the reasoning tier generates a long internal chain before it answers.

The practical upshot: a reasoning tier's price-per-token understates its real cost, because it spends far more tokens per answer. Reserve K2-thinking for problems where the chain-of-thought actually earns its keep, and never default to it for short, factual calls. The full token-by-token detail lives in the K2-thinking pricing micro.

Decision matrix: which tier, for which job, on which route

No single article in this cluster states this cross-topic synthesis, so here it is. Match the job to the tier, then the tier to a route.

The honest takeaway is that Kimi competes on long context and coding, not on being the absolute cheapest token. According to DeepSeek's pricing documentation, DeepSeek V4-Flash bills $0.14 input / $0.28 output, well under base K2's $0.57 / $2.30 via OpenRouter. According to DeepSeek's USD pricing details, those rates apply at the official api.deepseek.com endpoint, which is why our Kimi-via-OpenRouter versus DeepSeek-official comparison is not a perfectly matched routing test. If your task does not need Kimi's 256K window or its coding tune, a cheaper model probably wins, and we say so in the head-to-head micros below.

Every Kimi pricing question, mapped to its article

This hub is the canonical landing page. Each focused question has its own article. Here is the full cluster navigation.

According to Moonshot AI's documentation, international access to the Kimi platform is available through platform.moonshot.ai with USD-denominated plans distinct from the mainland RMB pricing, so US and international developers can reach the official endpoints directly. The access details, including which tiers are exposed, are covered in the US availability article.

How the routes differ, and why we disclose it

There is an asymmetry in this cluster's evidence that we want on the record. Every Kimi K2 price and measurement here is OpenRouter-routed, because the native Moonshot key in our environment only reaches the legacy moonshot-v1 endpoints, not the K2 family. The official Moonshot K2 prices are sourced research and carry a re-verification flag until we confirm them against a native K2 key. When we compare Kimi to DeepSeek, the DeepSeek number is from its official endpoint, so a Kimi-via-OpenRouter versus DeepSeek-official comparison is not a perfectly matched routing test, and we flag that wherever it appears.

That disclosure is the point of a neutral directory. A price you cannot trace to a route is a price you cannot budget against.

FAQ

What is the cheapest Kimi tier? The base K2 (K2-0711) generation is the cheapest, at $0.57 input / $2.30 output per million tokens measured via OpenRouter, with the official legacy K2 at near parity ($0.55 / $2.20). The reasoning and flagship tiers cost more.

Is the OpenRouter Kimi price cheaper than Moonshot's own? No. For base K2 the OpenRouter rate ($0.57 / $2.30) is a slight premium over the official legacy K2 ($0.55 / $2.20), roughly two cents per million. Pick a route for access and reliability, not for a price gap on base K2.

Does the OpenRouter "kimi-k2" alias mean the flagship K2.6? No. The moonshotai/kimi-k2 alias resolves to the legacy K2-0711 base generation. The flagship K2.6, at $0.95 / $4.00 with a 256K context according to Moonshot's pricing, is a separate and pricier tier you must request explicitly.

Why does K2-thinking cost so much more per answer? The reasoning tier generates a long internal chain before answering. On our test, the same prompt that cost base K2 60 output tokens cost K2-thinking 553 output tokens, about 9x the billed cost and more than double the latency, all measured via OpenRouter.

How does Kimi compare to DeepSeek on price? DeepSeek V4-Flash is cheaper, at $0.14 / $0.28 per million according to DeepSeek's docs, versus base K2's $0.57 / $2.30 via OpenRouter. Kimi competes on its 256K context and coding tune, not on being the cheapest token.

Can US developers use Kimi? Yes. According to Moonshot's documentation, platform.moonshot.ai offers USD-denominated international plans distinct from mainland RMB pricing, so US developers can reach the official endpoints. The full access detail is in the US availability article.

What context window does Kimi K2 offer? The flagship K2.6 and K2.7-Code tiers carry a 262,144-token (256K) context, Kimi's signature long-context strength, according to Moonshot pricing. The base K2 alias is a smaller, cheaper generation.

---

Author: Kevin Fan, Customer Success Manager at China LLM Directory, specializing in Chinese LLM ecosystem pricing and model benchmarking. Last verified: 2026-06-26.

<!-- METADATA { "title": "Kimi K2 (Moonshot) API Pricing Guide 2026", "slug": "kimi-api-pricing", "meta_description": "Base Kimi K2 is $0.57/$2.30 via OpenRouter, near parity with official legacy K2 $0.55/$2.20; flagship K2.6 is $0.95/$4.00 with 256K context. 2026-06.", "focus_keyword": "kimi api pricing", "secondary_keywords": ["kimi k2 pricing", "moonshot api pricing", "kimi k2.6 cost", "kimi k2 thinking price", "kimi api cost"], "tags": ["Kimi", "Moonshot", "API Pricing", "Long Context"], "category": "Pricing", "cluster_id": "kimi-api-pricing", "cluster_role": "hub", "evidence_file": "clients/china-llm-aggregator/articles/kimi-api-pricing-evidence.json", "needs_native_reverify": true, "verified_until": "2026-09-24", "author_name": "Kevin Fan", "author_title": "Customer Success Manager", "author_linkedin": "", "author_expertise": ["Chinese LLM ecosystem", "AI infrastructure pricing", "model benchmarking", "cross-border AI compliance"], "faq_pairs": [ {"q": "What is the cheapest Kimi tier?", "a": "The base K2 (K2-0711) generation is the cheapest, at $0.57 input / $2.30 output per million tokens measured via OpenRouter, with the official legacy K2 at near parity ($0.55 / $2.20). The reasoning and flagship tiers cost more."}, {"q": "Is the OpenRouter Kimi price cheaper than Moonshot's own?", "a": "No. For base K2 the OpenRouter rate ($0.57 / $2.30) is a slight premium over the official legacy K2 ($0.55 / $2.20), roughly two cents per million. Pick a route for access and reliability, not for a price gap on base K2."}, {"q": "Does the OpenRouter kimi-k2 alias mean the flagship K2.6?", "a": "No. The moonshotai/kimi-k2 alias resolves to the legacy K2-0711 base generation. The flagship K2.6, at $0.95 / $4.00 with a 256K context according to Moonshot pricing, is a separate and pricier tier you must request explicitly."}, {"q": "Why does K2-thinking cost so much more per answer?", "a": "The reasoning tier generates a long internal chain before answering. On our test, the same prompt that cost base K2 60 output tokens cost K2-thinking 553 output tokens, about 9x the billed cost and more than double the latency, all measured via OpenRouter."}, {"q": "How does Kimi compare to DeepSeek on price?", "a": "DeepSeek V4-Flash is cheaper, at $0.14 / $0.28 per million according to DeepSeek's docs, versus base K2's $0.57 / $2.30 via OpenRouter. Kimi competes on its 256K context and coding tune, not on being the cheapest token."}, {"q": "Can US developers use Kimi?", "a": "Yes. According to Moonshot's documentation, platform.moonshot.ai offers USD-denominated international plans distinct from mainland RMB pricing, so US developers can reach the official endpoints. The full access detail is in the US availability article."}, {"q": "What context window does Kimi K2 offer?", "a": "The flagship K2.6 and K2.7-Code tiers carry a 262,144-token (256K) context, Kimi's signature long-context strength, according to Moonshot pricing. The base K2 alias is a smaller, cheaper generation."} ], "external_links_used": [ {"url": "https://platform.moonshot.ai/", "source_name": "Moonshot AI Platform Pricing", "claim": "Kimi K2.6 flagship $0.95 in / $4.00 out, 256K context, $0.19 cached input"}, {"url": "https://platform.moonshot.ai/docs", "source_name": "Moonshot AI Documentation", "claim": "International USD plans via platform.moonshot.ai distinct from mainland RMB"}, {"url": "https://platform.moonshot.ai", "source_name": "Kimi/Moonshot Pricing", "claim": "Official legacy K2 $0.55/$2.20; OpenRouter base kimi-k2 ~2 cents/M higher"}, {"url": "https://api-docs.deepseek.com/quick_start/pricing/", "source_name": "DeepSeek API Docs - Pricing", "claim": "DeepSeek V4-Flash $0.14 in / $0.28 out per million"}, {"url": "https://api-docs.deepseek.com/quick_start/pricing-details-usd", "source_name": "DeepSeek API Docs - USD Pricing Details", "claim": "Official api.deepseek.com endpoint rates, basis for the routing-asymmetry disclosure"} ], "internal_links_used": [ {"url": "/blog/kimi-k2-pricing/", "anchor_text": "Kimi K2 pricing", "type": "micro"}, {"url": "/blog/kimi-k2-thinking-pricing/", "anchor_text": "Kimi K2-thinking pricing", "type": "micro"}, {"url": "/blog/kimi-k2-vs-deepseek-cost/", "anchor_text": "Kimi K2 vs DeepSeek cost", "type": "micro"}, {"url": "/blog/kimi-k2-vs-gpt-4o-cost/", "anchor_text": "Kimi K2 vs GPT-4o cost", "type": "micro"}, {"url": "/blog/cheapest-kimi-model/", "anchor_text": "Cheapest Kimi model", "type": "micro"}, {"url": "/blog/kimi-k2-coding-cost/", "anchor_text": "Kimi K2 coding cost", "type": "micro"}, {"url": "/blog/kimi-long-context-cost/", "anchor_text": "Kimi long-context cost", "type": "micro"}, {"url": "/blog/is-kimi-available-in-us/", "anchor_text": "Is Kimi available in US", "type": "micro"}, {"url": "/blog/kimi-api-cost-calculator/", "anchor_text": "Kimi API cost calculator", "type": "micro"} ], "first_hand_evidence": { "source": "kimi-api-pricing-evidence.json runs: kimi-k2_general, kimi-k2-thinking_general, kimi-k2.7-code_coding", "measured": "Base K2 60 out / 7.03s / $0.00015852; K2-thinking (kimi-k2-thinking-20251106) 553 out / 18.51s / $0.0014041 (~9x base on same prompt); K2.7-Code 201 out / 6.83s / $0.00083784. All routed via OpenRouter.", "captured": "2026-06-26" }, "images_status": "spec-only (not generated; FAL_API_KEY unset)", "images": [ {"position": "featured", "type": "generated", "prompt": "Clean editorial diagram of the Kimi K2 model family showing four tiers (base K2, K2-thinking, K2.6 flagship, K2.7-Code) as price-stacked bars annotated with input/output rates and route labels (OpenRouter vs official Moonshot). Deep blue and amber palette, minimal background, 16:9.", "alt": "Diagram of the Kimi K2 model family showing four tiers stacked by price with input and output rates and their routes labeled OpenRouter or official Moonshot"} ] } -->