# Kimi K3.1 Pro

> Kimi K3.1 Pro is the first model in Moonshot AI's K3.1 series, released on October 23, 2026. Built on Kimi K3's 3T-class Mixture-of-Experts architecture, it adds large-scale joint reinforcement learning across hundreds of domains, with targeted gains in coding, 3D creation (Blender, FreeCAD, Unreal Engine and Godot) and financial research. Moonshot reports 61.2% on Kimi Code Bench 3.0 Long-horizon (up from 55.95% on K3), 65.7% on Terminal-Bench 4.0 and 64.4 on the OfficeQA Pro v2 financial document benchmark; on Kimi Hallu Bench Hard, claim-level hallucination drops from 23.4% on K3 to 8.3%, and the share of responses containing at least one factual error falls to 42.4% of K3's rate.

Canonical: https://routemux.com/moonshot/kimi-k3.1-pro
Provider: Moonshot

## Key facts

| Field | Value |
| --- | --- |
| Model ID | moonshot/kimi-k3.1-pro |
| Official model ID | kimi-k3.1-pro |
| Context window | 1M |
| Max output | 1M |
| Input modalities | text, image |
| Output modalities | text |
| Capabilities | Vision, Reasoning, Tool Calling, Coding, Cache, Paid Cache Write |
| Upstream release | 2026-10-23 |
| Availability | available |

## Pricing

- RouteMux input: $1.6 per 1M tokens
- RouteMux output: $8 per 1M tokens
- Official input: $2 per 1M tokens
- Official output: $10 per 1M tokens
- Cache read: $0.16 per 1M tokens
- Cache write: $1.6 per 1M tokens
- Saving vs official: 20%

Billing is prepaid wallet, charged per successful request only.

## API access

Base URL: https://api.routemux.com

Supported protocols:

- openai_chat
- openai_responses
- anthropic_messages

## What is Kimi K3.1 Pro

Kimi K3.1 Pro is the first model in Moonshot AI's K3.1 series, released on October 23, 2026. Built on Kimi K3's 3T-class Mixture-of-Experts architecture, it adds large-scale joint reinforcement learning across hundreds of domains, with targeted gains in coding, 3D creation (Blender, FreeCAD, Unreal Engine and Godot) and financial research. Moonshot reports 61.2% on Kimi Code Bench 3.0 Long-horizon (up from 55.95% on K3), 65.7% on Terminal-Bench 4.0 and 64.4 on the OfficeQA Pro v2 financial document benchmark; on Kimi Hallu Bench Hard, claim-level hallucination drops from 23.4% on K3 to 8.3%, and the share of responses containing at least one factual error falls to 42.4% of K3's rate.

The context window is 1M tokens (1,048,576), and max_completion_tokens defaults to 131072 and can be raised up to 1048576. Vision is native, and on RouteMux it accepts text and image input. Reasoning is always on and cannot be disabled; you set its depth with the top-level reasoning_effort field, which accepts low, high and max and defaults to max. Moonshot's parameter rules still apply: temperature 1.0, top_p 0.95, n 1 and both penalties at 0 are fixed, and sending other values returns an error, so leave them out of the request; in multi-turn chats and tool calls, return the full assistant message unchanged, including reasoning_content; and tool_choice accepts auto, none and required. It is callable over OpenAI Chat Completions, OpenAI Responses and Anthropic Messages.

Against Kimi K3: both share the 1M context window, 1,048,576 output ceiling and reasoning_effort levels, while K3.1 Pro adds cross-domain joint reinforcement learning for stronger long-horizon coding, 3D modeling and financial analysis with fewer hallucinations. Stay on kimi-k3 only when existing prompts or evaluation baselines are pinned to the July 2026 K3 release. Moonshot has also announced that the K2 series (including kimi-k2.7-code) will retire at the end of December 2026 and recommends migrating to kimi-k3.1-pro.

On RouteMux, Kimi K3.1 Pro costs $1.6 per 1M input tokens (official $2) and $8 per 1M output tokens (official $10), with cached input at $0.16 per 1M, 20% below the official price.

## Key capabilities

- **Joint RL post-training** — Post-trained on Kimi K3 across hundreds of domains, reaching 61.2% on Kimi Code Bench 3.0 Long-horizon.
- **1M context, up to 1M output** — max_completion_tokens defaults to 131072 and goes up to 1048576.
- **Much lower hallucination rate** — Claim-level hallucination on Kimi Hallu Bench Hard drops from 23.4% on K3 to 8.3%.
- **Three effort levels & vision** — Reasoning is always on with reasoning_effort low, high or max (default max), plus native image input.
- **Three protocols** — OpenAI Chat, OpenAI Responses, and Anthropic Messages all route to this model.

## Use cases

- **Long-horizon coding & 3D creation** — Moonshot highlights full-stack engineering alongside 3D scene and asset generation in Blender, FreeCAD, Unreal Engine and Godot.
- **Financial research & filings** — Scores 64.4 on OfficeQA Pro v2 for financial document analysis, suited for long filings, research reports and data extraction.
- **Fact-critical deep research** — With claim-level hallucinations cut to 8.3% on Kimi Hallu Bench Hard, it fits research workflows that demand strict factual grounding.

## Other variants in this family

- Kimi K3 (`moonshot/kimi-k3`)

## Sources

- [Kimi API Platform — Kimi K3.1 Pro quickstart](https://platform.kimi.ai/docs/guide/kimi-k3-1-pro-quickstart)
- [Kimi API Platform — Model parameter reference](https://platform.kimi.ai/docs/api/models-overview)
- [Kimi API Platform — Chat pricing](https://platform.kimi.ai/docs/pricing/chat)