Kimi K3 Pricing: Plans & API Token Cost Calculator 2026
Five Kimi plans run from the free Adagio to Vivace at $199 a month, with annual billing dropping Vivace to $159, and the K3 API meters at $3.00 per 1M input tokens ($0.30 per 1M on a cache hit) and $15.00 per 1M output.
Kimi K3 plans and pricing
Kimi membership
Adagio
Free
- Kimi chat with the current model lineup
- Work with Docs, Sheets and Slides
- Deep Research and Websites Deploy to try
- Kimi Code at base (1x) credits
Moderato
Advanced Flow
- Everything in Adagio, with paid-tier throughput
- Agent multi-tasking
- Kimi Code available
- Slides visual mode with Nano Banana
- Dashboard with task widgets, scheduled tasks
- Plugins with financial and economic data
- Build and publish websites with database support
Allegretto
Pro Choice
- Everything in Moderato, doubled agent capacity
- Goal mode: agents work autonomously to completion
- Deploy your Kimi Claw in one click
- Slides visual mode with Nano Banana
- Scheduled tasks that run and push results
Allegro
1M-token chat unlocks herePremium Mode
- K3 Extra Long Chat Capacity: up to 1M tokens
- Four agent tasks running at once
- Swarm at 8 subagents, multiplied productivity
- Goal mode and one-click Kimi Claw deploy
- 100 projects and 50GB of storage
Vivace
Ultimate Boost
- The full Allegro feature set at maximum credits
- K3 Extra Long Chat Capacity: up to 1M tokens
- Highest agent and Kimi Code allowances sold
- For daily heavy agent and research workloads
- Kimi Enterprise exists separately for teams (unpriced)
Developer API: the full model catalog
kimi-k3
Flagship- 2.8T-parameter MoE with native image and video input
- Reasoning always on (reasoning_effort: max only)
- Only model with tool_choice: required and dynamic tools
- Flat rates: no context-length pricing bands
- No Batch discount for K3 at launch
kimi-k2.7-code
Coding- Dedicated coding model, text, image and video input
- The lineup's most stable structured output
- HighSpeed variant: ~180 tok/s at exactly 2x the price
- Covered by the 60% Batch discount
kimi-k2.6
Cost saver- General multimodal, thinking and non-thinking modes
- The vendor's own pick for routing cost-sensitive traffic
- Second most-used LLM on OpenRouter in May 2026
- Covered by the 60% Batch discount
kimi-k2.5
Sunset Aug 31- Cheapest current-gen model on the platform
- Closes to new users around August 31 (expected 2026)
- Covered by the 60% Batch discount
- Migration path: k2.6 at $0.95/$4.00
moonshot-v1
Legacy- Previous-generation family, priced per context size
- Vision-preview variants match non-vision prices
- Only listed for accounts still migrating off it
| Model | Cache-hit input | Cache-miss input | Output | Context |
|---|---|---|---|---|
| kimi-k3 | $0.30 | $3.00 | $15.00 | 1,048,576 |
| kimi-k2.7-code | $0.19 | $0.95 | $4.00 | 262,144 |
| kimi-k2.7-code-highspeed | $0.38 | $1.90 | $8.00 | 262,144 |
| kimi-k2.6 | $0.16 | $0.95 | $4.00 | 262,144 |
| kimi-k2.5 (sunset Aug 31) | $0.10 | $0.60 | $3.00 | 262,144 |
| moonshot-v1-8k (legacy) | $0.20 flat | — | $2.00 | 8,192 |
| moonshot-v1-32k (legacy) | $1.00 flat | — | $3.00 | 32,768 |
| moonshot-v1-128k (legacy) | $2.00 flat | — | $5.00 | 131,072 |
| The same 50K-token prompt | What one call costs | Why |
|---|---|---|
| Cold, first time seen | $0.15 | 50,000 tokens at the $3.00 per 1M cache-miss rate |
| Warm, on a cache hit | $0.015 | The same tokens at $0.30 per 1M, billed automatically |
| Cache-hit rate | Effective input per 1M | Where it comes from |
|---|---|---|
| 0% | $3.00 | Every call cold, novel prompts each time |
| 50% | $1.65 | Half your context reused |
| 90% | $0.57 | Vendor's claimed hit rate in coding workloads |
| 93% | $0.489 | OpenRouter's observed launch-week average |
| Tier | Recharge to reach | Concurrency | Requests/min | Tokens/min |
|---|---|---|---|---|
| Tier0 | $1 | 1 | 3 | 500,000 |
| Tier1 | $10 | 50 | 200 | 2,000,000 |
| Tier2 | $20 | 100 | 500 | 3,000,000 |
| Tier3 | $100 | 200 | 5,000 | 3,000,000 |
| Tier4 | $1,000 | 400 | 5,000 | 4,000,000 |
| Tier5 | $3,000 | 1,000 | 10,000 | 5,000,000 |
| Model | Cache-hit input | Cache-miss input | Output |
|---|---|---|---|
| kimi-k3 | ¥2.00 | ¥20.00 | ¥100.00 |
| Output tokens on the same eval | Volume | What it means for a bill |
|---|---|---|
| Kimi K3 | ~130M | Close to double a terse model; output dominates most bills |
| Field average | ~63M | The baseline the $15.00 per 1M rate assumes |
Prices shown verbatim from kimi.com and platform.kimi.ai, verified July 17, 2026. Membership cards open the Kimi membership page, API cards the Moonshot developer platform. China-platform rates (¥2 cache-hit / ¥20 cache-miss / ¥100 output per 1M for K3) are independent price points, not conversions. All links go through our tracked, protected redirect.
Kimi K3 pricing: the quick answer
Kimi runs on five consumer plans plus one pay-as-you-go API, and the plans are the place to start. The ladder goes from the free Adagio (one agent task, two projects, 500MB of storage) through Moderato at $19 a month and Allegretto at $39 to Allegro at $99, the tier where the K3 Extra Long Chat Capacity of up to 1M tokens switches on, and tops out at Vivace at $199 with the highest agent and Kimi Code credit allowances. Annual billing trims every paid tier, Vivace dropping from $199 to $159 a month, and the vendor has flagged this lineup as changing soon, with Kimi and Kimi Code benefits due to be separated. For developers, the K3 API meters at $3.00 per 1M input tokens on a cache miss, $0.30 per 1M on a cache hit, and $15.00 per 1M output at a 1M-token context; caching is automatic, a Launch Top-Up Rebate of 10-30% runs through August 12, 2026, and output is where API bills actually grow because reasoning is always on and the model is verbose.
- AdagioFree
- API Pay-as-you-go$3/1M input tokens
- Moderato$19/mo
- Allegretto$39/mo
- Allegro$99/mo
- Vivace$199/mo
At $3/mo to start, Kimi K3 sits 62% below the $7.99/mo median across 11 large language models tools we track.
Kimi K3 cost calculator
Kimi K3 Hidden Costs & Pitfalls
Kimi K3 has no API subscription. The bill is set by three things: your cache-hit rate on input, how verbose the always-on reasoning makes your output, and which model you route to.
Kimi K3 pricing, read against its live plans and category
Positioning
Kimi K3 is priced to make the Opus tier look expensive. The API runs $3.00 per 1M input tokens on a cache miss, $0.30 on a cache hit, and $15.00 per 1M output, all at the full 1M context with no length bands. Artificial Analysis scores its Intelligence Index at 57 (fourth of 189), which is ahead of Claude Opus 4.8 at 56, so you are buying frontier-adjacent quality at a fraction of the frontier price. On the consumer side, five membership tiers run from the free Adagio to Vivace at $199 a month; the tier that actually matters is Allegro at $99, because that is where the K3 Extra Long Chat Capacity of up to 1M tokens switches on. Annual billing trims each paid tier, up to $480 a year on Vivace.
Cost drivers
- 1Always-on reasoning means every call pays for thinking tokens, and AA flags K3 as very verbose (130M output tokens on its eval against a 63M average), so real output cost runs high.
- 2The `$web_search` tool bills $0.005 per successful call plus the search-result tokens as input, and the vendor currently marks it as being updated and not recommended.
- 3The 60% Batch discount does not cover K3 at launch, only k2.7-code, k2.6, and k2.5.
- 4Prices exclude tax; the China entity invoices 6% VAT. Recharges are non-refundable.
Watch-outs
Moonshot publishes no compliance certifications (no SOC 2, ISO 27001, or HIPAA anywhere in the docs), and by default API content may be used to improve the models unless you have an enterprise or written agreement.
Strengths
Automatic prompt caching is the genuine differentiator: a cache hit drops input from $3.00 to $0.30 per 1M, a 90% discount, with no cache ID, TTL, or storage fee to manage. OpenRouter's launch-week traffic averaged an effective input of $0.488 per 1M at a 93% cache-hit rate. The full budgeting model behind these numbers, with three modeled monthly bills, lives in the Kimi K3 cost guide.
- 1M context as standard, no context-length bands
- Open weights (Modified MIT) expected around July 27, 2026, so self-hosting is a real path
- Launch Top-Up Rebate through August 12, 2026: a one-time top-up of $20 or more earns a 10-30% voucher, capped at $4,000 and valid 90 days
Editor’s take
Developers should run K3 on the pay-as-you-go API and treat cache-hit rate and reasoning verbosity as the two dials that set the bill. If you need published compliance paperwork or a no-training-by-default guarantee, an Opus-tier provider is the safer buy despite the higher token price. Whether the model itself earns the spend is the full Kimi K3 review's question, company file included.
Oleh KemFounder & Lead AnalystKimi K3 price history
Cheaper Large Language Models tools
Frequently asked questions
Sources & verification
| Source | What was checked | Last checked |
|---|---|---|
| Kimi K3 rate card (platform docs) | The $3.00 / $0.30 / $15.00 per 1M rates and cache mechanics | July 17, 2026 |
| Launch Top-Up Rebate terms | The 10-30% voucher bands, cap and window, verbatim | July 17, 2026 |
| Kimi membership pricing | The five consumer tiers and annual billing | July 17, 2026 |
| OpenRouter: Kimi K3 | Observed 93% cache-hit rate behind the effective-price table | July 17, 2026 |
| Official Website | Official vendor website | — |
Every fact on this Kimi K3 pricing page is tied to a named source and a verification date. Freshness-sensitive figures trace to the sources above; verify against the vendor before relying on them.
Explore Kimi K3
Every page on Kimi K3 in one place, you are on pricing.
Snapshot, score and verdict
How to get API access, limits, SDKs and what it costs
You are here
Hidden fees, discounts and how to negotiate
Compared and ranked vs peers
Price and feature change history
Browse the full Large Language Models category