Kimi K3.1 Pro Highspeed

Kimi K3.1 Pro Highspeed is Moonshot’s separately priced Highspeed K3.1 Pro API variant. Its API identity and billing are separate from Kimi Code subscriptions and the promised weight release.

Kimi K3.1 Pro explainer
Code/API, variant and weight boundaries.
Standard Pro API
Compare the separately priced variant.
Kimi K3
The distinct predecessor.

Language ModelCoding AssistantVisionAgent Workflows

Moonshot/Kimi hosted API. Code plan access is separate; downloadable weights/license unverified.

$4 input/$20 output; cache writes $4 (5m)/$8 (1h), hits $0.20 per million tokens. API balance/rate limits and Kimi Code subscriptions are separate.

Scoped Terms

🚀Function Overview

Kimi K3.1 Pro Highspeed: 1,048,576 context and always-on thinking. API $4 input/$20 output; cache writes $4 (5m)/$8 (1h), hits $0.20 per million tokens.

Key Features

API context 1,048,576; max_completion_tokens defaults to 131,072, up to 1,048,576 within the total context budget.
Always-on thinking; top-level reasoning_effort accepts low, high or max, API default max; Code default high.
Text and visual inputs; visual input uses base64 or ms:// file IDs, not public image URLs.
Omit fixed temperature=1.0, top_p=0.95, n=1 and zero presence/frequency penalties rather than treating them as tunable.
Return complete assistant messages unchanged in multi-turn/tool conversations.
API token rates: $4 input/$20 output; cache writes $4 (5m)/$8 (1h), hits $0.20 per million tokens.
Kimi Code IDs omit kimi- prefix; subscription entitlement, context access and quota differ from API prepaid balance.
Moonshot describes about 2–3x speed and similar capability; this is vendor-reported, not a measured GetLLMs latency result.

Use Cases

•Supervised coding and document analysis.
•Visual reasoning and long-running tool workflows.

⚙️Input Parameters

model

string

Use kimi-k3.1-pro-highspeed on the general paid API.

messages

array

Chat-completion messages; preserve the full returned assistant message on subsequent turns.

reasoning_effort

string

low, high or max (default max); thinking cannot be disabled.

max_completion_tokens

integer

Default 131072, up to 1048576 within the shared context budget.

💡Usage Examples

Example 1

Input Parameters

{
  "model": "kimi-k3.1-pro-highspeed",
  "messages": [
    {
      "role": "user",
      "content": "Inspect the supplied code and propose one bounded change with a verification plan."
    }
  ],
  "reasoning_effort": "high"
}

Output Results

Illustrative request or command only. GetLLMs has not executed this example or verified model output.

Quick Actions

Technical Specifications

Hardware Type: Moonshot/Kimi hosted API. Code plan access is separate; downloadable weights/license unverified.
Commercial Use: Hosted service use follows the selected provider terms. This record does not establish a downloadable-weight license.
Pricing: $4 input/$20 output; cache writes $4 (5m)/$8 (1h), hits $0.20 per million tokens. API balance/rate limits and Kimi Code subscriptions are separate.