Models
What Is Kimi K3.1 Pro? Moonshot’s New Model, Price and Benchmarks
Kimi K3.1 Pro is Moonshot AI’s new flagship, out October 23, 2026. What it is, what it costs, how it scores against Claude Opus 5.5, and when the weights arrive.
Kimi K3.1 Pro is the new flagship AI model from Moonshot AI, the Beijing company behind the Kimi chatbot. It went live on October 23, 2026, and within hours it was the top AI story on Hacker News and in tech news feeds.
If you saw the name and wondered what it is, whether it is free, and why people keep comparing it to Claude, this page answers all of that in plain words.
What is Kimi K3.1 Pro?
Kimi K3.1 Pro (people also type it as “Kimi K3.1”, “Kimi 3.1 Pro” or “kimi-k3.1-pro”) is a large language model made by Moonshot AI. Moonshot calls it the first model in its new K3.1 series.
It is not a brand-new design. It is Kimi K3 with a lot more training on top. Moonshot describes it as a “3T-class” open-weight model. The Decoder reports about 2.8 trillion total parameters, with roughly 104 billion active for any one answer. Think of a huge library where only the right few shelves get opened for each question. That keeps it faster and cheaper than its size suggests.
It takes text and images as input and has a 1-million-token context window. That is enough to read a long contract, a big codebase or a stack of reports in one go.
Kimi K3.1 Pro release date
Moonshot released Kimi K3.1 Pro on October 23, 2026. Some outlets list October 24, which is simply the date in China at the time of launch. The Highspeed version came out the same day. The open weights are a separate step: Moonshot says the full model weights will be released by November 4, 2026.
Kimi K3.1 Pro price: is it free?
On the API, you pay per token. Here are the launch prices, per million tokens:
| Model | Input | Output | Cached input |
|---|---|---|---|
| Kimi K3.1 Pro | $2.00 | $10.00 | $0.20 |
| Kimi K3.1 Pro Highspeed | $4.00 | $20.00 | $0.20 |
| Kimi K3 (previous) | $3.00 | $15.00 | $0.30 |
So K3.1 Pro is about 33% cheaper than K3. Highspeed costs twice as much and, per Moonshot, runs two to three times faster with the same skills. Moonshot’s docs say the flagship unlocks on the API after a top-up of at least $1.
Is it free in the Kimi app? Not clearly. Moonshot’s plan notes tie K3.1 Pro to paid tiers (Plus or above, with Highspeed and the full 1M context on Pro or above). A free tier for this model has not been confirmed. Once the weights are out, anyone with serious hardware will be able to run it themselves at no license fee, but the license terms have not been published yet.
Kimi K3.1 Pro benchmarks vs Claude Opus 5.5 and GPT-6 Astra
The headline claim is coding. Most numbers below come from Moonshot itself and have not been checked by outside testers, except where noted.
| Test | Kimi K3.1 Pro | Claude Opus 5.5 | GPT-6 Astra |
|---|---|---|---|
| Terminal-Bench 4.0 (coding agents) | 65.7% | 66.4% | 58.2% |
| BenchCAD, with tools | 92.7% | 96.2% | 95.9% |
| Artificial Analysis Intelligence Index (independent) | 55, 4th place | 58, 1st place | Below Kimi |
The independent score matters most. Artificial Analysis puts K3.1 Pro fourth overall, behind Claude Opus 5.5, Claude Sonnet 5.5 and Meta’s Muse Wonder Full, which puts it ahead of GPT-6 Astra. Kimi K3 scores 44 on the current version of the same index, so this is a big jump for one update.
Moonshot also says K3.1 Pro makes up fewer facts. On its own hard stress test, the share of false claims fell from 23.4% for K3 to 8.3%. Moonshot itself warns that this test does not reflect everyday error rates.
Why Moonshot cut its hacking skills
This is the unusual part. Because the weights will be public, and open models cannot be recalled once downloaded, Moonshot says it removed cyberattack-related data from training while keeping normal coding skills.
On Moonshot’s in-house test of exploiting real software flaws, a version trained without those cuts solved 90.5% of tasks. The released K3.1 Pro solved 81%. K3 solved 33.3%. So the cut was real but small, and the model is still far stronger here than K3. Moonshot says its cyber skills remain behind the best closed models. That trade-off is worth watching, since AI-found bugs like PHP2Shell are already making headlines.
How to try Kimi K3.1 Pro
- Kimi apps: Kimi.ai on the web and the Kimi apps for iPhone, Android and HarmonyOS, on a plan that includes it.
- Kimi Work: the desktop app, version 3.1.0 or later, on Windows and Apple silicon Macs.
- Kimi Code: pick the model with the
/modelcommand. Highspeed uses about twice the quota. - API: model IDs
kimi-k3.1-proandkimi-k3.1-pro-highspeed. Moonshot offers Zero Data Retention for API customers. - Open weights: due by November 4, 2026, for teams that want to host it themselves.
What this means for you
For most people, Kimi K3.1 Pro is one more sign of how fast the gap is closing. A model that is cheap, nearly as good as the top closed models at coding, and soon free to download makes strong AI a commodity. The hard part is no longer finding a smart model. It is turning that intelligence into finished work, like a plan, a draft or a summary you can act on.
That is what an agent app is for. Instead of signing up for a new chatbot every week, you describe the job and a team of agents splits it up and hands back the result. You can learn how that works in our guide to what agentic AI is.
Quick answers
What is Kimi K3.1 Pro? It is Moonshot AI’s newest flagship AI model, a post-trained upgrade of Kimi K3 with a 1-million-token context window, released October 23, 2026.
What is the Kimi K3.1 release date? Kimi K3.1 Pro and K3.1 Pro Highspeed launched October 23, 2026 (October 24 in China). The open weights are due by November 4, 2026.
Is Kimi K3.1 Pro free? Not confirmed. API use costs $2 per million input tokens and $10 per million output tokens, and Moonshot ties the model to paid Kimi plans. Downloadable weights are promised for November 4.
Is Kimi K3.1 Pro better than Claude Opus 5.5? Not overall. Moonshot’s own figures put it just behind Opus 5.5 on Terminal-Bench 4.0 (65.7% vs 66.4%), and the independent Artificial Analysis index ranks it fourth, behind Opus 5.5.
What is Kimi K3.1 Pro Highspeed? The same model served faster. Moonshot says it runs two to three times faster, at double the price: $4 input and $20 output per million tokens.
Is Kimi K3.1 Pro open source? It will be open-weight. Moonshot says the full weights arrive by November 4, 2026. The license has not been published yet.
More new AI names, explained simply, in our agentic AI glossary.
Sources: Moonshot AI: Kimi K3.1 Pro quickstart · The Decoder · The Next Web · Unite.AI · Vercel AI Gateway (Highspeed pricing) · Hacker News discussion
