Kimi K3.1 Pro: Three Points Away, With Weights Yet To Arrive
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Kimi K3.1 Pro: Three Points Away, With Weights Yet To Arrive on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Moonshot AI released Kimi K3.1 Pro in its apps and API on October 23. Artificial Analysis scored it 55.1, three points behind the leading model in its comparison, but the model’s weights are still promised for November 4 and no licence has been published.

Moonshot AI released Kimi K3.1 Pro in its apps, coding product and API on October 23, with the model scoring 55.1 on Artificial Analysis’s Intelligence Index. That puts it fourth in the cited comparison, three points below the highest-scoring model, but the weights are not yet available: Moonshot has promised to publish them by November 4, and has not announced a licence for the release.

Artificial Analysis lists K3.1 Pro as a mixture-of-experts model with 2.8 trillion total parameters and 104 billion active parameters. It accepts text and images and has a reported one-million-token context window. The service began appearing on October 23 across the Kimi apps, Kimi Work, Kimi Code and the Kimi API.

In the Artificial Analysis comparison dated October 24, K3.1 Pro scored 55.1. Claude Opus 5.5 scored 57.6, Claude Sonnet 5.5 scored 56.0, and Muse Wonder Full scored 55.8. Artificial Analysis says K3.1 Pro would lead open-weight models once its weights are published; that ranking is conditional because the weights have not been released. Its index comparison places the previously leading open-weight model, Xiaomi’s MiMo-V2.6-Pro, at 46.3.

Artificial Analysis lists K3.1 Pro at $2 per million input tokens and $10 per million output tokens, with an estimated $2.11 per Intelligence Index task. Its estimate for Claude Opus 5.5 is $5.98 per task. These are evaluation-cost figures, not a guarantee of what every user’s workload will cost. The comparison used each model at its highest reasoning setting, according to the source material.

At a glance
updateWhen: Launched October 23, 2026; weights prom…
The developmentMoonshot AI launched Kimi K3.1 Pro, which scored 55.1 on Artificial Analysis’s Intelligence Index, while its weights remain unreleased.
Thorsten Meyer AI
Moonshot AI · 23 Oct 2026

Infographic 01 · Claimed vs. measured

Moonshot says “matches the strongest.” Independent tests say: close.

The KPIs at the top are independent measurements. Vendor numbers and independent numbers come from different harnesses, so they are shown side by side, never mixed.

55Intelligence Index score: 4th of all models
3 ptsbehind the leader, Claude Opus 5.5 (58)
67%Terminal-Bench 4.0: highest AA has measured
$2.11per index task, vs. $5.98 for Opus 5.5 (max)

Moonshot reports

vendor claims own harness, partly internal tests

BenchmarkK3.1 ProComparison
Terminal-Bench 4.065.7%vs. Claude Opus 5.5: 66.4%
BenchCAD with tools92.7%vs. Claude Opus 5.5: 96.2%
OfficeQA Pro v264.4vs. Claude Opus 5.5: 62.2
Kimi Hallu Bench Hard internal8.3%vs. Kimi K3: 23.4% (about −65%)

Artificial Analysis measured

independent highest reasoning setting per model

MeasurementK3.1 ProComparison
Terminal-Bench 4.067%vs. Claude Opus 5.5 (max): 60%
GDPval-AA1,846 Elovs. Claude Opus 5.5: 1,866
Humanity’s Last Exam52%vs. Claude Opus 5.5: 61%
AA-Omniscience hallucination rate44%vs. Kimi K3: 53% · Gemini 4 Argon: 15%

The 65% hallucination drop holds on Moonshot’s own stress test. On an independent knowledge test, the rate fell from 53% to 44%.Moonshot: the stress test rates “do not represent overall error rates in daily use.”

Sources: Moonshot AI, Kimi K3.1 Pro tech blog (23 Oct 2026) (left card, vendor charts); Artificial Analysis launch article (23 Oct 2026) and Artificial Analysis model page (data as of 24 Oct 2026) (KPIs and right card).
Powered by Thorsten Meyer AI

A Near-Frontier Score, With a Caveat

The result matters because it narrows the measured performance gap between an unreleased open-weight model and leading proprietary systems. If Moonshot publishes usable weights under a licence that permits the intended uses, developers and organisations could have more control over where they run the model and how they integrate it. That possibility remains conditional: neither the weights nor the K3.1 licence is available yet.

For organisations comparing hosted services, the listed token prices and task-cost estimate suggest K3.1 Pro may be less expensive than Opus 5.5 for the benchmark’s measured work. But benchmark rankings do not establish performance on every company task, and hosted pricing does not answer the cost of operating the model independently. Given its 2.8-trillion total parameters, self-hosting would require substantial infrastructure; the source describes it as a data-centre-scale undertaking.

The reported gains also vary by task. Artificial Analysis recorded 67% on Terminal-Bench 4.0, compared with 13% for Kimi K3 in its run, while the knowledge benchmark Humanity’s Last Exam rose from 47% to 52%. Those results suggest a particularly strong showing on agentic coding tasks, not uniform leadership across all categories.

Products Worth Considering

Apple MacBook Neo 13-inch Laptop with A18 Pro Chip, 8GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID – Blush

Apple MacBook Neo 13-inch Laptop with A18 Pro Chip, 8GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID – Blush

Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID (Midnight)

Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID (Midnight)

As an affiliate, we earn on qualifying purchases.

From Kimi K3 to K3.1 Pro

K3.1 Pro is presented as a more heavily post-trained version of Kimi K3. The source’s Artificial Analysis figures show Kimi K3 at 43.6 on the Intelligence Index, compared with 55.1 for K3.1 Pro. The same comparison lists K3 as an open-weight model under the Kimi K3 License; it does not establish what terms will govern K3.1 Pro.

Moonshot’s launch materials also describe projects in game engines, parametric 3D modelling and finance. Those examples and much of the supporting evaluation come from the company itself. For instance, Moonshot reports 92.7% on BenchCAD with tools, compared with 96.2% for Claude Opus 5.5 and 95.9% for GPT-6 Astra. These should be treated as vendor-reported results, distinct from Artificial Analysis’s independent evaluations.

Hallucination figures illustrate why the distinction matters. Moonshot says its internal Kimi Hallu Bench Hard test showed a claim-level hallucination rate falling from 23.4% for K3 to 8.3% for K3.1 Pro, a reduction of about 65%. Artificial Analysis measured a decrease from 53% to 44% on its AA-Omniscience evaluation. The tests use different methods, so their percentages are not directly interchangeable. Moonshot also cautions that its internal test was deliberately designed to provoke errors and does not represent typical daily-use error rates.

“Kimi K3.1 Pro is its most capable “3T-class open-weight model.””

— Moonshot AI, in its launch materials

Weights, Licence and Real-World Results

The weights have not been published, and the promised November 4 date is a future target rather than an accomplished release. Moonshot has also not published the K3.1 licence in the material cited here. Until both are available, developers cannot confirm the terms for use, modification or redistribution, or whether the released model will match the evaluated version.

It is also unclear how the model’s benchmark performance will translate to routine work. Artificial Analysis provides independent scores on selected evaluations, while Moonshot’s game, CAD and finance examples rely largely on company demonstrations and internal tests. The supplied information does not establish independent replication of those demos or provide enough detail to compare total deployment costs across different workloads.

Even the reported hallucination reductions need careful interpretation: the vendor’s internal test and Artificial Analysis’s evaluation use different measures and produce different rates. Neither number, on its own, establishes an everyday error rate across general use.

The November 4 Release Test

The next milestone is whether Moonshot publishes K3.1 Pro’s weights by its stated November 4 deadline, and what licence accompanies them. The release would let independent developers examine the model and test whether its performance can be reproduced outside Moonshot’s services. Until then, access is through the company’s listed apps, products and API.

After publication, the licence terms, model files and independent evaluations will help clarify whether organisations can run the model for their intended purposes and what hardware and operating costs that would involve. Readers should also watch for updates to Artificial Analysis’s rankings and for third-party tests of the vendor’s coding, CAD, finance and hallucination claims. No such post-release results are confirmed in the source material.

Key Questions

What is Kimi K3.1 Pro?

Kimi K3.1 Pro is a Moonshot AI model described as a more heavily post-trained version of Kimi K3. Artificial Analysis lists 2.8 trillion total parameters, 104 billion active parameters, text and image input, and a one-million-token context window.

How did it rank against other models?

It scored 55.1 on Artificial Analysis’s Intelligence Index in the comparison dated October 24, 2026. That placed it fourth among the models listed, behind Claude Opus 5.5 at 57.6, Claude Sonnet 5.5 at 56.0 and Muse Wonder Full at 55.8.

Can developers download the weights now?

No. Moonshot has promised the weights by November 4, 2026, but they had not been published in the source material. The licence for K3.1 Pro also had not been announced.

What does Kimi K3.1 Pro cost?

Artificial Analysis lists API prices of $2 per million input tokens and $10 per million output tokens. It estimates $2.11 per Intelligence Index task. Actual costs depend on usage, and running the weights independently would add infrastructure and operating expenses.

Are Moonshot’s claims about hallucinations independently confirmed?

Not in the same form. Moonshot reports a fall from 23.4% to 8.3% on its internal Kimi Hallu Bench Hard test. Artificial Analysis reports a decrease from 53% to 44% on its separate AA-Omniscience evaluation. The tests differ, and Moonshot says its own test does not represent general daily-use error rates.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

Meta Stock

Meta’s stock declined sharply after its latest earnings report, reflecting investor concerns over revenue growth and future outlook.

Decoding Qwen3.8-Max’s AI Performance Metrics

Alibaba officially released detailed benchmarks and specs for Qwen3.8-Max, confirming its 2.4T parameters and top-tier performance, with open weights arriving soon.

7 Best PC Tablets for Prime Day Deals in 2026

Discover the best PC tablets on Prime Day 2026, including the Samsung Galaxy Tab S9, Surface Pro 11, and iPad 9th Gen, with detailed analysis and buying tips.

Protect Yourself And Others With Aftermarket Fatigue Monitoring

New phone-based fatigue alerts aim to warn long-commute drivers of drowsiness, filling safety gaps in older vehicles without built-in tech.