🎁 K3 Launch Promo: 10–30% bonus credits on all prepaid top-ups · Claim by Aug 11, 2026 →
Try Kimi Free →
INCLUDES KIMI K3 PRICING

Simple, Transparent
Kimi AI Pricing

Five plans from free to $199/month. K3 is now the flagship - accessible in the app across all tiers, with full 1M context unlocking at Allegretto. API billing always separate from membership.

Monthly
Annual Save up to $480/yr
🚀 Launch Promotion - Ends Aug 11, 2026

K3 is live. Get 10–30% bonus credits on all prepaid API top-ups at platform.kimi.ai through August 11, 2026. One-time top-up of ¥199 activates API access.

+10–30% bonus credits

All plans support 256K context minimum (1M ctx on Allegretto+). API billing is always separate from membership. Prices in USD. Taxes not included. Quotas subject to change - verify at platform.kimi.ai.

// FEATURE COMPARISON

What's Included in Each Plan

Feature AdagioFree Moderato$19/mo Allegretto$39/mo Allegro$99/mo Vivace$199/mo
// API PRICING - ALL MODELS

API Pricing by Model

API billing is completely separate from your membership plan - token-based, pay-as-you-go. One API key unlocks all available models. Get your key at platform.kimi.ai.

Kimi K3 NEW · #1 Frontend Code
2.8T params · 1M ctx · Vision + Video · kimi-k3
$3.00/1M in $15.00/1M out $0.30/1M cached · $0.015/search
Kimi K2.7 Code HighSpeed ⚡ 260 tok/s
1T MoE · 256K ctx · Vision · kimi-k2.7-code-highspeed
$1.90/1M in $8.00/1M out $0.19/1M cached
Kimi K2.7 Code Kimi Code default
1T MoE · 256K ctx · Vision · −30% reasoning tokens · kimi-k2.7-code
$0.95/1M in $4.00/1M out $0.19/1M cached
Kimi K2.6 GA · Agent Swarm
1T MoE · 262K ctx · Vision · 300 agents · SWE-bench 80.2% · kimi-k2.6
$0.55/1M in $2.65/1M out $0.19/1M cached
Kimi K2.5 ⚠ Retiring Aug 31
1T MoE · 256K ctx · MoonViT 400M · 100 agents · kimi-k2.5 (check retirement notice)
$0.60/1M in $3.00/1M out $0.10–0.16/1M cached

Quick start

  • Get key at platform.kimi.ai (minimum ¥199 top-up)
  • Base URL: https://api.moonshot.ai/v1
  • OpenAI SDK, LangChain, LlamaIndex, Anthropic SDK all work
  • Swap base_url and model from any existing client
  • Also on openrouter.ai/moonshotai/kimi-k3 at same rates

⚠ Model retirements - Aug 31, 2026

Moonshot has issued a retirement notice for kimi-k2.5 and moonshot-v1 on August 31, 2026. Migration to kimi-k2.7-code (routine coding) or kimi-k3 (max quality) is recommended. K2.7 Code is not in the retirement notice.

Python · OpenAI SDK · All Models
from openai import OpenAI client = OpenAI( api_key="YOUR_KIMI_KEY", base_url="https://api.moonshot.ai/v1" ) # K3 - max quality, always thinking, 1M ctx r = client.chat.completions.create( model="kimi-k3", max_tokens=32768, messages=[{"role":"user", "content":"Audit this 800K-line codebase..."}] ) # K2.7 HighSpeed - fast iterative coding (Beta) r = client.chat.completions.create( model="kimi-k2.7-code-highspeed", max_tokens=32768, messages=[{"role":"user", "content":"Fix the race condition in auth.ts"}] ) # K2.6 - best cost / quality for agents r = client.chat.completions.create( model="kimi-k2.6", extra_body={"thinking": False}, # instant mode messages=[{"role":"user", "content":"Generate a unit test suite..."}] ) # Model selection guide: # K3 → max quality, long ctx, vision+video # K2.7 HS → fast iterations, Beta access required # K2.7 Code → default; best cost/quality balance # K2.6 → agent swarm tasks, cheapest frontier

Cost per task - real-world estimates

  • K3 typical task: ~$0.94 · vs Opus 4.8 ~$1.80 · vs Fable 5 ~$3.20
  • K3 cache-heavy: 90%+ cache hit rates on coding → effective ~$0.30/1M in
  • K2.7 Code: best for high-frequency iteration loops
  • K2.6: cheapest option for agent swarm + multi-step tasks
  • Always-on thinking on K3 + K2.7 means reasoning tokens bill as output at full rate
// MODEL ACCESS BY PLAN

Which Model on Which Plan

All tiers get access to Kimi K3 in the app. What changes by plan is the context window available, the agent credit quota, and Kimi Code multiplier.

Kimi K3
Released Jul 16, 2026 · 2.8T params
Adagio (Free)Basic access · 256K ctx
Moderato ($19)K3 + 256K ctx
Allegretto ($39)Full 1M context ✦
Allegro ($99)Full 1M context ✦
Vivace ($199)Full 1M · Priority queue ✦
Kimi K2.7 Code HighSpeed
Released Jun 15, 2026 · 260 tok/s · Beta
Adagio – ModeratoNot available
Allegretto+ / BetaBeta Program access
API (all keys)kimi-k2.7-code-highspeed
💻
Kimi K2.7 Code
Released Jun 12, 2026 · Kimi Code default
Adagio (Free)Limited access
Moderato ($19)Kimi Code 1× credits
Allegretto ($39)Kimi Code 5× credits
Allegro ($99)Kimi Code 15× credits
Vivace ($199)Kimi Code 30× credits
🤖
Kimi K2.6 - Agent Swarm
GA · 300 agents · 4,000 steps
Adagio – ModeratoNo Agent Swarm
Allegretto ($39)50 uses · max 4 agents
Allegro ($99)120 uses · max 6 agents
Vivace ($199)240 uses · max 8 agents
📊
Professional Data
World Bank · Market data · Academic
Adagio (Free)200 requests/month
Moderato ($19)2,000 requests/month
Allegretto ($39)5,000 requests/month
Allegro ($99)12,000 requests/month
Vivace ($199)24,000 requests/month
🦞
Kimi Claw (OpenClaw)
24/7 cloud agent · 5,000+ skills
Adagio – ModeratoNot available
Allegretto ($39)Claw access · 40GB storage
Allegro ($99)Priority Claw scheduling
Vivace ($199)Maximum capacity + SLA
// COST CALCULATOR

Estimate Your API Cost

Cost per request-
Cost per day-
Cost per month (30d)-
Cost per year-
Select a model and adjust parameters to see your estimate.
// QUICK ANSWER

How Much Does Kimi AI Cost?

Short answer: free to start, $19–$199/month for membership, and token-based API from $0.30/1M (cached) to $15/1M. Here is the full picture in one place.

Option 1 - App Membership
$0 – $199/mo

Five tiers: Adagio (free) → Moderato ($19) → Allegretto ($39) → Allegro ($99) → Vivace ($199). Covers the web app, mobile, Kimi Code CLI, agent credits, Slides, and all tools. Does NOT include API calls.

Option 2 - Pay-per-Token API
$0.30 – $15/1M tok

Billed per token at platform.kimi.ai. K3: $3/$15 per 1M. K2.7: $0.95/$4.00. K2.6: $0.55/$2.65. Cache hits drop input to $0.19–$0.30/1M. Separate from membership.

Option 3 - Self-Host (Open Weights)
$0 + GPU infra

K2.6, K2.7 Code, K3 weights available under Modified MIT. K3 open weights ship July 27, 2026. Deploy on A100-class clusters with vLLM, SGLang, or TensorRT-LLM.

Use case Best option Typical monthly cost
Casual chat and researchAdagio (Free)$0
Daily professional useModerato $19 + API$25–$60
Developer building with APIAPI only (K2.6 or K2.7)$5–$200 (usage-based)
Agent swarm workflowsAllegretto $39 + API$60–$200
K3 - frontier quality tasksAPI kimi-k3~$0.94/task
High-volume AI infrastructureVivace $199 + API + self-host$500–$5,000+
// ALL KIMI AI PLANS

All Kimi AI Plans, Side by Side

A single reference table covering every plan dimension - monthly and annual rates, context access, model credits, tools, and quotas.

Dimension AdagioFree Moderato$19/mo Allegretto$39/mo Allegro$99/mo Vivace$199/mo
// KIMI K3 PRICING

Kimi K3 Pricing - Full Breakdown

Released July 16, 2026. The world's largest open-weight model at 2.8T parameters - priced at the same tier as Claude Sonnet, but delivering frontier-level coding performance.

$0.30CACHED INPUT / 1M
$3.00FRESH INPUT / 1M
$15.00OUTPUT / 1M
Model ID: kimi-k3 & kimi-k3-swarm-max · Also on OpenRouter: moonshotai/kimi-k3 · Web search: $0.015/call · Flat across full 1M context
Cost per typical task
~$0.94
vs Opus 4.8 ~$1.80 · vs Fable 5 ~$3.20
Cache hit rate (coding)
90%+
Mooncake serving · effective ~$0.30/1M in
Long-context surcharge
None
Flat $3/$15 across full 1,048,576 tokens
Cheaper than Fable 5 by
~3.3×
Fable 5 output ~$50/1M vs K3 $15/1M
⚠ Watch: Always-on thinking = billed output

K3 has no non-thinking mode. Every response includes a reasoning trace billed at the full $15.00/1M output rate. A chatty thinking trace (common on complex tasks) can cost more than the visible answer. K3 is also the only SKU - there is no cheaper "instant" K3 variant.

K3 cost estimate · Python
# Estimate K3 cost for a coding session in_tokens = 80_000 # 80K input (large repo) out_tokens = 12_000 # 12K output (code + thinking) cache_hit = True # 90%+ hit on coding tasks inp_rate = 0.30 if cache_hit else 3.00 out_rate = 15.00 cost = (in_tokens/1_000_000) * inp_rate + \ (out_tokens/1_000_000) * out_rate # cache hit: $0.024 + $0.18 = $0.204 # cache miss: $0.24 + $0.18 = $0.420 print(f"Cost: ${cost:.3f}")

K3 access by membership plan

Adagio (Free)K3 in app · 256K context
Moderato ($19)K3 + priority · 256K context
Allegretto ($39)K3 + full 1M context ✦
Allegro ($99)K3 + full 1M context ✦
Vivace ($199)K3 + 1M + dedicated queue ✦
Direct API (any plan)Full 1M always · $3/$15/1M
// KIMI K2.7 CODE PRICING

Kimi K2.7 Code & HighSpeed Pricing

Released June 12–15, 2026. The default model for Kimi Code CLI. Two speed tiers: standard and HighSpeed (Beta, 5–6× faster model output at 2× the token price).

Kimi K2.7 Code
Standard · Kimi Code Default
Released Jun 12, 2026 · Model ID: kimi-k2.7-code
$0.19CACHED IN/1M
$0.95FRESH IN/1M
$4.00OUTPUT/1M
  • 1T MoE · 256K context · Vision via MoonViT 400M
  • Mandatory thinking mode · −30% reasoning tokens vs K2.6
  • MCP Mark Verified 81.1 · beats Claude Opus 4.8 (76.4)
  • Best cost/quality ratio for iterative coding workflows
  • NOT in the Aug 31 retirement notice (K2.5 is)
⚡ Kimi K2.7 Code HighSpeed - Beta
260 tok/s · 6× Faster
Released Jun 15, 2026 · Model ID: kimi-k2.7-code-highspeed
$0.19CACHED IN/1M
$1.90FRESH IN/1M
$8.00OUTPUT/1M
  • Same model as K2.7 Code - optimized serving infrastructure
  • Up to 260 tok/s peak · ~180 tok/s on median coding inputs
  • 2× the token cost of standard K2.7 · choose when speed matters
  • Speed applies to model output only - tool call time is separate
  • Available: Kimi Code Beta · Allegretto+ app · Direct API (all)
// KIMI K2.6 PRICING

Kimi K2.6 Pricing - Best Value for Agents

General availability since April 20, 2026. The cheapest frontier-class model in the Kimi lineup - and the one to use when cost matters for agent swarm and multi-step workflows.

$0.19CACHED IN/1M
$0.55FRESH IN/1M
$2.65OUTPUT/1M
Model ID: kimi-k2.6 · 1T MoE · 262K context · Vision · SWE-bench 80.2%
5.5×
cheaper output than K3
$2.65 vs $15.00/1M
300
parallel agents
4,000 steps in Agent Swarm

When to choose K2.6 over K3

  • Running hundreds of agent steps per task (cost adds up fast at K3 rates)
  • Agent Swarm workflows - K2.6 is the swarm backbone model
  • High-volume API calls where quality difference doesn't justify 5× cost
  • Tasks needing 262K context (slightly larger than K2.7/K3's 256K)
  • Non-thinking / instant mode needed (K2.6 supports switchable thinking)
// K2.6 vs K2.7 Code vs K3 - at a glance
Metric K2.6 K2.7 K3
Input $/1M$0.55$0.95$3.00
Output $/1M$2.65$4.00$15.00
Cached $/1M$0.19$0.19$0.30
Context window262K256K1M
Parameters1T1T2.8T
Thinking modeSwitchableMandatoryMandatory
Video input--
Agent Swarm300 agents-Swarm Max
Open weights✓ Live✓ LiveJul 27, 2026
// TOKEN BILLING EXPLAINED

Token Billing Explained

Tokens are the unit of all Kimi API billing. Everything - input, output, reasoning traces, system prompts, conversation history, tool calls, and image descriptions - is measured in tokens.

What gets billed as tokens?

INPUT

System prompt + conversation history + your message + any uploaded images/video descriptions. Billed once per request. Cache hits dramatically reduce this cost.

OUTPUT

Visible answer + thinking trace (on K2.7 Code and K3). Always-on thinking models produce thinking tokens at the full output rate - this is where K3 costs can surprise you.

CACHED

When Mooncake serving recognizes repeated content (your system prompt, repo context, document), the input cost drops to $0.19/1M (K2.x) or $0.30/1M (K3). Cache hit rates on coding tasks exceed 90%.

TOOLS

Tool invocations (file reads, shell commands, web scraping) consume tokens when results are returned to the model. Web search costs an additional flat $0.015/call.

Token size reference

1 token ≈ 4 characters (English)
1,000 tokens ≈ 750 words
1M tokens ≈ 750,000 words
K3 1M ctx ≈ 800K lines of code
Image ≈ 1K–4K tokens depending on size
Video frame ≈ similar to an image

Billing flow - one API request

1. System prompt (2K tokens)cached → $0.0006
+
2. Repo context (50K tokens)cached → $0.015
+
3. New user message (500 tokens)fresh → $0.0015
+
4. Thinking trace (4K tokens)output → $0.060
+
5. Visible answer (2K tokens)output → $0.030
+
6. Web search call (×1)flat → $0.015
Total (K3, with cache)~$0.122 / request
// MEMBERSHIP VS API

Membership vs API: Which Is Right for You?

These are separate systems that serve different use cases. Most power users need both. Here is how to decide.

Membership Plan ($0–$199/mo)

For the Kimi app experience

Choose membership when you want:
  • Chat UI, mobile app, and Kimi Work desktop agent
  • Kimi Code CLI with included credits (no per-token billing)
  • Agent Swarm, Slides, Docs, Sheets, Websites tools
  • Professional Data quotas and Kimi Claw cloud agent
  • Predictable monthly cost regardless of usage volume
Best for: Individual professionals, content creators, researchers, developers who use the Kimi interface daily and want predictable billing.
API Pay-per-Token

For programmatic access

Choose API when you want:
  • Embed Kimi into your own app, product, or pipeline
  • Fine-grained control: choose model, context, temperature
  • Pay only for what you use - no fixed monthly floor
  • Access the latest models including K3 at full 1M context
  • Scale from zero to millions of tokens per day
Best for: Developers building AI products, companies running automated workflows, teams with variable or high-volume usage patterns.
💡 Most power users combine both: A Moderato or Allegretto membership for the app + Kimi Code CLI + agent credits, plus an API account for programmatic access and custom integrations. The $19/month membership pays for itself in Kimi Code credits alone for most developers.
// COST DRIVERS

What Drives Kimi API Costs Up

Token prices are simple - what makes bills unpredictable are the multipliers. Here are the seven biggest cost drivers across Kimi models.

🧠

Always-on thinking traces

K3 and K2.7 Code always produce a reasoning trace before answering. Thinking tokens bill as output at the full output rate ($15/1M for K3, $4/1M for K2.7). On complex tasks the thinking trace can be 2–5× longer than the visible answer.

📜

Long conversation history

Every turn sends the full conversation history as input. A 10-turn session where each exchange is 5K tokens means turn 10 pays for 50K+ input tokens. Cache hits reduce this but don't eliminate it - new user messages always count as fresh input.

🔍

Web search calls

Each web search invocation costs $0.015 flat, plus the tokens consumed reading results. In a Deep Research session running 100+ searches, that's $1.50 in search calls alone before counting input/output tokens.

🤖

Agent loops and tool calls

Each agent step is a separate API call - with the full accumulated context as input. A 300-step agent session at 10K average context per step = 3M input tokens before any output. Caching helps only for repeated context.

🖼️

High-resolution images

Images are tokenized before processing. A high-res screenshot can consume 2K–4K tokens of input. In a multimodal coding session with 10 UI screenshots, that's 20K–40K tokens of image-only input per request.

📄

Lack of prompt caching

If your system prompt or document context changes every request, you pay full fresh-input rates every time. Pin your system prompt and load documents once; let Mooncake's cache handle the rest. This alone can cut input costs by 80–90%.

Choosing K3 for every task

K3 output costs 5.5× more per token than K2.6 ($15 vs $2.65). For tasks like summarization, classification, or quick code edits where K2.6 or K2.7 produces equivalent results, using K3 is pure cost inflation with no quality benefit.

// PLAN SELECTOR

Which Plan for Which User?

Skip the comparison tables. Find your profile and get a direct recommendation.

// KIMI AI VS COMPETITORS

Kimi AI vs Other AI Tools - Pricing & Value

How Kimi's plans and API rates compare to ChatGPT, Claude, Gemini, Perplexity, and Copilot - across subscription price, API cost, open weights, context, and agent capability.

Dimension Kimi AIMoonshot AI ChatGPTOpenAI ClaudeAnthropic GeminiGoogle PerplexityPerplexity AI
// AGENT API PRICING

Kimi Agent API Pricing

There is no separate "agent" pricing tier. Agent workflows are billed at the same per-token rates - but agents multiply token consumption in ways that matter for cost planning.

How agent pricing works

Each agent step = 1 API call

A 300-step Agent Swarm session makes 300+ API calls. Each call sends the full accumulated task context as input. No special "agent" discount - regular token rates apply.

Tool calls add tokens

File reads, shell output, web scraping results - all returned to the model as additional input tokens. A file-heavy coding agent reading 50 files across a session adds tens of thousands of input tokens.

Web search: $0.015/call

Every web search invocation by the agent costs $0.015 flat plus tokens for reading search results. Plan for $1–$3 in search fees per Deep Research session.

💡 Cost-saving tip: For long agent sessions, use K2.6 ($0.55/$2.65) for agent steps and call K3 only for the final synthesis step. A 200-step agent session on K2.6 then 1 K3 synthesis call costs ~5–10× less than running all 201 steps on K3.

Agent API cost estimate - 50-step coding agent

50 steps × 8K avg input (K2.6)~$0.22
50 steps × 2K avg output (K2.6)~$0.27
10 web search calls$0.15
Total (K2.6 agent)~$0.64
Same session on K3 (50 steps)~$7.50
K3 = 11.7× more expensive for this agent session

Agent Swarm (membership)

Agent Swarm capacity (50–240 swarm uses, 4–8 agents) is included in Allegretto+ membership plans. Using Agent Swarm via the Kimi app does not charge additional API tokens from your API account - it runs against membership credits. Only direct API calls to kimi-k2.6 for agent orchestration are token-billed.

// MULTIMODAL API PRICING

Kimi Multimodal API Pricing

Images and video are billed in tokens - no separate multimodal surcharge. The cost depends on the model and media resolution, not a per-image flat fee.

🖼️

Images

Images are tokenized based on resolution. A typical UI screenshot at 1280×800 consumes approximately 1,500–3,000 input tokens. Available on K2.5 (via MoonViT), K2.6, K2.7 Code, and K3.

Rough token counts:
512×512 px ≈ 400–800 tokens
1280×800 px ≈ 1,500–2,500 tokens
2560×1600 px ≈ 4,000–6,000 tokens
Cost (K3) = tokens × $3.00/1M input
🎬

Video (K3 only)

K3 is the only Kimi model supporting native video input. Frames are extracted and each is tokenized as an image. A 60-second clip at 1 frame/second = ~60 image-tokens. Resolution determines per-frame token count.

Estimate:
30s clip, 1 fps, 720p ≈ 45K–90K tokens
Cost (K3) = ~$0.14–$0.27 input
Available on: K3 only
Model ID: kimi-k3
📊

Model support matrix

K3 (kimi-k3)Text + Image + Video
K2.7 CodeText + Image
K2.6Text + Image
K2.5Text + Image (retiring)
K2 InstructText only
// LONG-CONTEXT API PRICING

Kimi Long-Context API Pricing

K3's 1M token context is the largest in the Kimi lineup. Critically, Moonshot charges flat rates with no long-context surcharge - unlike some competitors that charge premium rates above 200K tokens.

Flat pricing across all context lengths
K3 - 0 to 1,048,576 tokens
No context-length premium
$3.00/1M
K2.7 Code - 0 to 256K tokens
Flat rate
$0.95/1M
K2.6 - 0 to 262K tokens
Flat rate · largest window in K2 family
$0.55/1M
Practical K3 1M-context cost: Sending 900K cached input tokens (full codebase, already in Mooncake) + 100K fresh input (new context) + 20K output:
= 900K × $0.30/1M + 100K × $3.00/1M + 20K × $15/1M
= $0.27 + $0.30 + $0.30 = $0.87 total

KDA: Why K3's 1M context is usable

Standard transformer attention scales quadratically with context length (O(n²)). At 1M tokens, this is computationally prohibitive. Kimi Delta Attention (KDA) replaces most attention with a linear recurrence, inserting sparse "delta" full-attention passes only where needed.

6.3×Faster decode at 1M ctx
~$0/1MLong-context surcharge

Context access by membership plan

Adagio + Moderato256K max in app
Allegretto ($39/mo)Full 1M context ✦
Allegro ($99/mo)Full 1M context ✦
Vivace ($199/mo)Full 1M + priority ✦
Direct APIFull 1M always (set max_tokens)
// WEB SEARCH API PRICING

Kimi Web Search API Pricing

Web search is billed as a flat tool-call fee per invocation, on top of the regular token charges for reading and processing search results.

Per search call
$0.015
per invocation · flat rate
+ tokens for results

How it's billed

$0.015 flat per web_search tool call
+ tokens for search results returned to the model
+ tokens for any web pages read (if model browses further)
Applicable to all models with tool use: K3, K2.7, K2.6

Cost at scale

10 searches$0.15
100 searches (Deep Research)$1.50
1,000 searches / day$15.00/day
10,000 searches / day$150.00/day

When search is invoked

  • Model decides to search based on your prompt (agentic mode)
  • Deep Research mode - up to 100+ searches per session
  • You can pass web_search as a tool in the tools array

When to suppress search

  • Don't include web_search in tools when not needed
  • For high-volume batch processing where search isn't required
  • For code generation or analysis tasks that don't need live web data
// FAQ

Pricing Questions Answered

Start Free. Scale When You're Ready.

Free Adagio plan includes Kimi K3, Deep Research, Slides Adaptive mode, and 200 Professional Data requests per month - no credit card.