XCoffee
Back to blog

Kimi K3.1 Pro API: Pricing, 1M Context, How to Call It

Oct 28, 2026XCoffee Team

Kimi K3.1 Pro API pricing checked against Moonshot: $2/$10 per 1M tokens, 1M context, Highspeed, how to call it, plus Flint as a free coding fallback.

Kimi K3.1 Pro API: Pricing, 1M Context, How to Call It

The Kimi K3.1 Pro API launched on October 23, 2026. Moonshot AI's new flagship has a 1M-token context, costs $2 / $10 per 1M tokens, and is on both the Kimi API platform and OpenRouter. Below are the prices, the context limits and code to call it. At the end is a free coding fallback, Recursive's Flint, which arrived on OpenRouter two days earlier.

Sources, checked October 29, 2026: Moonshot's K3.1 Pro quickstart, Kimi's price list, and OpenRouter's public model list (/api/v1/models). The USD prices match on OpenRouter and in Moonshot's docs.

Kimi K3.1 Pro API at a glance

Kimi K3.1 Pro
Provider Moonshot AI
Kimi API model ID kimi-k3.1-pro (also kimi-k3.1-pro-highspeed)
OpenRouter ID moonshotai/kimi-k3.1-pro
What it is Multimodal reasoning model, post-trained successor to Kimi K3. Weights promised by Nov 4, 2026
Context window 1,048,576 tokens
Max output (OpenRouter) 943,718 tokens
Input types text, image, video
Input price $2.00 / 1M tokens
Output price $10.00 / 1M tokens
Cached input read $0.20 / 1M tokens
Cache write (Kimi API) $2.00 (5-min TTL) / $4.00 (1-hour TTL) per 1M
Batch variant not listed on OpenRouter

Moonshot positions it for long-horizon coding and agent work. Pricing is flat across the whole 1M context, with no long-context tier. On Moonshot's own CNY price list, K3.1 Pro (¥13 in / ¥68 out) costs less than Kimi K3 (¥20 / ¥100).

Kimi K3.1 Pro Highspeed and access rules

  • Highspeed (kimi-k3.1-pro-highspeed) is 2–3× faster and costs $4 / $20 per 1M tokens. Cached input stays $0.20, and cache writes are $4 / $8.
  • On the Kimi API, K3.1 Pro unlocks after a minimum $1 top-up. Your total top-up sets your rate limits.
  • Moonshot says the full weights will be released by November 4, 2026. Until then it is API-only. OpenRouter already calls it open-weight.
  • Moonshot fixes temperature, top_p and penalties, so leave them out of requests. Its built-in web search is "not recommended for production" for now.

Free fallback: Recursive's Flint

Flint (free)
Provider Recursive
OpenRouter ID recursive/flint-v1.0:free
Architecture Open-weight MoE, ~2.5B active / 40B total parameters
Built by Recursive's "Eureka Machine" automated research system, for local agentic coding
Context window 262,144 tokens on OpenRouter (Recursive says up to 512K locally, best within 256K)
Max output 235,929 tokens
Input text only
Price $0 input / $0 output
Tools supported

Free OpenRouter models carry rate limits, and free tiers can end with little notice (MiniMax M3.1 Flash's did on October 22). OpenRouter does not publish an end date for Flint's free tier. Flint's weights are also on Hugging Face, so you can run it locally. For more free models, see our free LLM API roundup. With 2.5B active parameters, expect it to handle simple edits and tool calls, not hard multi-file refactors.

How to call the Kimi K3.1 Pro API

From Moonshot directly (OpenAI-compatible):

import os
from openai import OpenAI
client = OpenAI(api_key=os.environ["MOONSHOT_API_KEY"], base_url="https://api.moonshot.ai/v1")
r = client.chat.completions.create(
    model="kimi-k3.1-pro",
    messages=[{"role": "user", "content": "Write a unit test for this function: ..."}],
)
print(r.choices[0].message.content)

Through OpenRouter:

from openai import OpenAI
client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key="YOUR_OPENROUTER_KEY")
r = client.chat.completions.create(
    model="moonshotai/kimi-k3.1-pro",      # or "recursive/flint-v1.0:free"
    messages=[{"role": "user", "content": "Write a unit test for this function: ..."}],
)
print(r.choices[0].message.content)

Kimi K3.1 Pro pricing example: one agent day

An agent that sends 200 requests × 60K input + 3K output tokens (no caching):

  • Kimi K3.1 Pro: 12M × $2 + 0.6M × $10 = $30 a day
  • Flint (free): $0, if you stay within free-tier rate limits

Prompt caching ($0.20 per 1M cached reads) can cut the Kimi figure a lot when the same repo context is re-sent.

A flat-price option for coding agents

XCoffee plan details updated November 12, 2026.

XCoffee gives you one key with three groups: xcoffee-coding, xcoffee-fast and xcoffee. The relay picks the model inside the group with automatic failover; you cannot choose a specific model. As of October 29, 2026, the XCoffee homepage lists kimi-k3.1-pro among the published models in the xcoffee-coding group. A given request can still be served by any model in that group, so don't treat it as a Kimi K3.1 Pro endpoint.

  • 30-day passes have plan-specific prices, daily credit allowances and rolling hourly caps. See the current plans. Daily credits reset at UTC 00:00.
  • OpenAI (https://api.xcoff.ee/v1), Anthropic (https://api.xcoff.ee) and Gemini formats
  • Concurrency and request-rate caps depend on your plan; your account shows its actual limits. Provider-hosted tools are not included.

If you need Kimi K3.1 Pro specifically, use it directly. If you need an agent to keep working all day without a per-token bill, a flat key is simpler. Claude Code and Codex setup: tstack guide. More free models: Free LLM API options. Comparing OpenAI's new model: GPT-6.1 Terra API. Full setup: guide.

Kimi K3.1 Pro API FAQ

How much does the Kimi K3.1 Pro API cost?

$2 per 1M input tokens, $10 per 1M output and $0.20 per 1M cached reads, on both Moonshot's API and OpenRouter. Highspeed costs $4 / $20.

What is Kimi K3.1 Pro's context window?

1,048,576 tokens (1M), priced flat with no long-context tier.

Is Kimi K3.1 Pro open-weight?

Not yet. Moonshot says the full weights will be released by November 4, 2026. OpenRouter already describes it as open-weight.

Does Kimi K3.1 Pro have a batch version?

None is listed on OpenRouter as of October 29, 2026.

What is Flint?

An open-weight MoE model from Recursive (40B total, about 2.5B active) built for fast local coding. It is free on OpenRouter with a 262K context.

Is Flint really free?

The :free variant is $0 on OpenRouter, with rate limits. No end date is published.

Can I use Kimi K3.1 Pro through XCoffee?

It is listed in the xcoffee-coding group, but routing is automatic, so you can't force it.

What is the Kimi K3.1 Pro API model name?

kimi-k3.1-pro on Moonshot's API (base URL https://api.moonshot.ai/v1), and moonshotai/kimi-k3.1-pro on OpenRouter.