Grok 4.7
Step 5 Preview
Kimi K2.8 Preview
- input
- $0.2375
- cached
- $0.19
- output
- $8.7856
- Context
- 1Mtokens
- Released
- 2w ago
- Size
- 2.8Tparameters
Cost/1M tokens
Moonshot AI's Kimi K3.1 Pro is the first model in its K3.1 series and was post-trained from Kimi K3 using more post-training compute and large-scale reinforcement learning across many downstream domains. The maker describes it as its most capable 3T-class open-weight model to date. It is built for coding, 3D creation, financial research, and agentic tasks.
- Multimodal model
- Open weights · 2.8T
- Accepts images, video
- 1M context
- Reasoning
- Tools
- Structured output
How it compares
Score vs. costArena Text
Each model at its best score · Dashed: the best score at each cost
- Moonshot AI
- Google DeepMind
- Meta
- Z.ai
- Anthropic
- Xiaomi
- OpenAI
- xAI
Level
FlagshipThe lab's most capable generally available line, priced for hard tasks.
Flagship models from other labs
Vs. your stack
Add the models you use to see how this one compares.
Sign inAdditional source evidence
- Source model
- moonshotai/kimi-k3.1-pro
- Source family
- kimi-k3
- Last updated
- 2026-10-23
- Model context
- 1M
- Model output limit
- 131K
- Open weights
- Yes
- File attachments
- Yes
- Reasoning
- Yes
- Tool calling
- Yes
- Temperature control
- No
Complete source record (JSON)
{
"id": "moonshotai/kimi-k3.1-pro",
"name": "Kimi K3.1 Pro",
"description": "Multimodal Kimi model with a 1M-token context window and always-on reasoning, for coding, 3D creation and financial research",
"family": "kimi-k3",
"last_updated": "2026-10-23",
"attachment": true,
"reasoning": true,
"tool_call": true,
"temperature": false,
"modalities": {
"input": [
"text",
"image",
"video"
],
"output": [
"text"
]
},
"limit": {
"context": 1048576,
"output": 131072
},
"release_date": "2026-10-23",
"open_weights": true
}Frequently asked questions
Running locally
By our estimate, Kimi K3.1 Pro needs about 1,961 GB at Q4, 3,436.4 GB at Q8 and 6,468.1 GB at FP16 of VRAM. That covers the weights plus headroom for a short context, and longer contexts need more.
No, by our estimates Kimi K3.1 Pro does not fit fully on a 24 GB GPU even at Q4. It needs more VRAM or offloading to system memory.
Vertical Tabs records support for Kimi K3.1 Pro in Ollama.
API and pricing
Kimi K3.1 Pro is listed at $2.00 per 1M input tokens and $10.00 per 1M output tokens. Input prices across 22 providers range from $0.304 to $4.00 per 1M tokens.
Relace has the lowest blended price for Kimi K3.1 Pro (3:1 input to output), at $0.304 per 1M input tokens and $8.1953 per 1M output tokens.
Yes, cached input for Kimi K3.1 Pro costs $0.20 per 1M tokens, compared with $2.00 for uncached input, a 90% discount.
A request with 10,000 input tokens and 1,000 output tokens costs about $0.0112 with Kimi K3.1 Pro at the cheapest provider, Relace.
19 of 22 tracked Kimi K3.1 Pro endpoints had 99% uptime or better over the last day. The best uptime was 100.0%, shared by DeepInfra, DigitalOcean, InferenceNet and 7 more.
Privacy and data
Relace, Wafer, Sail Research, Morph, AkashML and 10 more offer zero data retention for Kimi K3.1 Pro, so prompts are not stored after the request.
Yes, Kimi K3.1 Pro has open weights, so you can run it on your own hardware and keep prompts on your own machines.
Capabilities and specs
Kimi K3.1 Pro has a 1M token context window. It can write up to 944K output tokens per response.
Yes, Kimi K3.1 Pro accepts images and video as input alongside text.
Yes, Kimi K3.1 Pro supports tool calling (function calling), so it can request calls to tools you define.
Yes, Kimi K3.1 Pro supports structured output, so responses can be constrained to JSON.
Yes, Kimi K3.1 Pro has a reasoning mode. Effort levels are max, high, low and none, with max as the default. 22 of 22 tracked providers expose it.
Quality and benchmarks
Kimi K3.1 Pro ranks #11 on Arena WebDev, #13 on Agent Arena and #24 on Arena Text, using Arena's published ranks.
Mostly yes: on 3 shared benchmarks, Kimi K3.1 Pro is ahead of Kimi K3 on 2, behind on 0 and level on 1. It leads on Agent Arena and Arena WebDev.
Similar models from other labs include Gemini 4 Carbon (Google DeepMind), MiMo-V2.6-Pro-UltraSpeed (Xiaomi) and Gemini 4 Argon (Google DeepMind). They are the closest matches by kind, price and release date.
Gemini 4 CarbonKimi K3.1 Pro vs Gemini 4 CarbonMiMo-V2.6-Pro-UltraSpeedKimi K3.1 Pro vs MiMo-V2.6-Pro-UltraSpeedGemini 4 ArgonKimi K3.1 Pro vs Gemini 4 Argon
Licensing and access
Kimi K3.1 Pro has open weights, so you can download and run it yourself.
The Kimi K3.1 Pro weights are on Hugging Face at moonshotai/Kimi-K3.1-Pro.
Basics
Kimi K3.1 Pro was made by Moonshot AI. It was released on October 23, 2026.
Yes, Kimi K3.1 Pro has open weights, so you can run it on your own hardware without API fees. Paid API access is listed at $2.00 per 1M input tokens and $10.00 per 1M output tokens.
Generated from Vertical Tabs data, updated November 11, 2026.
Discussion
No comments yet
Be the first to share your thoughts.