Try Chat Free Docs Pricing AI Analysis Repo Insights Free Tools About Blog
Sign in Get started free
< Back to blog
Research Velona Team ·16 November 2026 ·5 min read

Kimi K3 in India: Price in ₹, API Access and How It Compares With Claude and GPT (2026 Review)

Kimi K3 costs ₹304.19 per 1M input tokens and ₹1,622 per 1M output tokens on Velona. The newer Kimi K3.1 Pro costs ₹202.79 and ₹1,014.[7] Moonshot AI's own list price for K3 is $3 input and $15 output per 1M tokens, with cached input at $0.30.[1][2] Both models can be called from India with a UPI top-up and an OpenAI compatible API.

Key findings

Scope and method

This review draws on Moonshot's published prices as reported by four price and model guides, one independent benchmark summary and the Velona catalogue on 14 November 2026.[1][2][4][5][7] Benchmark figures are labelled by who reported them. We ran no tests of our own for this article. A chat reply is counted as 1,000 input and 500 output tokens, and a coding agent step as 30,000 input and 1,000 output tokens.

What Kimi K3 is

PropertyKimi K3Source
DeveloperMoonshot AI[2]
API launch16 July 2026[2]
Open weights27 July 2026, under the Kimi K3 License[2]
Size2.8 trillion parameters, 104 billion active[5]
Context window1,048,576 tokens, flat price at any length[1][2]
InputsText, image and video[5]
Reasoning levelsLow, high and max[2][5]

Kimi price in India in rupees

Model and Velona IDInput ₹ per 1MOutput ₹ per 1MOne chat replyOne agent step
Kimi K3.1 Pro
moonshotai/kimi-k3.1-pro
₹202.79₹1,014₹0.71₹7.10
Kimi K3
moonshotai/kimi-k3
₹304.19₹1,622₹1.12₹10.75
Kimi K2.6
moonshotai/kimi-k2.6
₹96.33₹405.58₹0.30₹3.30

Source: Velona catalogue, 14 November 2026.[7] One price guide reports that Moonshot closed the K2 series to new users on its own platform on 31 October 2026.[4] Check the model page before you build on K2.6.

How it compares on price

ModelInput ₹ per 1MOutput ₹ per 1MOne agent step
DeepSeek V4.1 Flash₹30.42₹121.67₹1.03
Kimi K3.1 Pro₹202.79₹1,014₹7.10
Claude Sonnet 5.5₹202.79₹1,014₹7.10
GPT-6.1 Sol₹202.79₹1,014₹7.10
Kimi K3₹304.19₹1,622₹10.75
Claude Opus 5.5₹405.58₹2,028₹14.20

The common idea that every Chinese model is very cheap does not hold for K3. One reviewer called it the priciest model any Chinese lab had shipped at launch.[3] The low cost option in this family today is K3.1 Pro, and for bulk work a model such as DeepSeek V4.1 Flash costs far less per step.

What the benchmarks show

The figures below are Moonshot's own reported evaluations, summarised by an independent guide.[5] Vendor tests are useful as a signal and should not be read as a neutral ranking.

Moonshot-reported testKimi K3Claude Fable 5GPT-5.6 Sol
GDPval-AA v2 Elo1,6681,7601,748
AA-Briefcase Elo1,5481,5831,495
JobBench52.957.446.5
BrowseComp91.288.090.4

An independent index gives a second view. One guide reports that Claude Opus 5 scored 51 against 44 for Kimi K3 on the Artificial Analysis Intelligence Index, while K3 cost about $2.00 per index task against $5.86.[6] In short, K3 is close to the frontier on several tests, behind on others, and lower in cost per task than Opus.

Where caching matters

First request with a long prompt
   input billed at full price        (Moonshot list: $3.00 per 1M)
        |
        v
Later requests that reuse the same prefix
   input billed at the cached rate   (Moonshot list: $0.30 per 1M)

Moonshot lists cached input at one tenth of the normal input price.[1] Coding agents resend the same files many times, so a warm cache changes the bill more than the headline price does. The agent step costs in this article assume no caching, so they are upper bounds.

How to call Kimi from India with UPI

  1. Create an account at velona.in and top up with UPI from ₹10.
  2. Create an API key and set a budget on it.
  3. Call the model with the OpenAI SDK.
from openai import OpenAI

client = OpenAI(api_key="YOUR_VELONA_KEY", base_url="https://velona.in/v1")
reply = client.chat.completions.create(
    model="moonshotai/kimi-k3.1-pro",
    messages=[{"role": "user", "content": "Review this function for edge cases: def div(a, b): return a / b"}],
    max_tokens=400,
)
print(reply.choices[0].message.content)
print(reply.usage)

Limitations

Frequently asked questions

What is the Kimi K3 price in India?

On Velona, ₹304.19 per 1M input tokens and ₹1,622 per 1M output tokens. Kimi K3.1 Pro is ₹202.79 and ₹1,014.[7]

Is Kimi K3 open source?

Its weights are public under the Kimi K3 License, released on 27 July 2026.[2] At 2.8 trillion parameters it is not practical to run on a personal computer.

Is Kimi K3 cheaper than Claude?

K3 lists below Claude Opus 5.5 and above Claude Sonnet 5.5. Kimi K3.1 Pro matches Sonnet 5.5 at $2 and $10 per 1M tokens.[4][7]

Can I pay for the Kimi API with UPI?

Yes, through a rupee wallet on Velona, with a one-time UPI top-up.

Disclosure: Velona publishes this review and sells access to these models. Prices were checked on 14 November 2026.

References

  1. eesel AI: Kimi K3 pricing, API cost and app tiers
  2. Morph: Kimi K3 specifications, launch dates and pricing
  3. Techsy: Kimi K3 review
  4. BenchLM: Kimi API pricing, November 2026
  5. Kingy AI: Kimi K3 benchmarks, pricing and evidence against named rivals
  6. Codersera: Kimi K3 complete guide, index comparison with Claude Opus 5
  7. Velona: live model pricing in INR