🔍 Read the full analysis: What Makes Kimi K3.1 Pro’s Frontier-Near Score Stand Out? on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Artificial Analysis gave Moonshot AI’s Kimi K3.1 Pro an Intelligence Index score of 55.1, placing it fourth in its October 24 comparison and three points behind the leader. The model is available through Moonshot’s services, but its weights are only promised by November 4, its licence has not been published, and vendor demonstrations are not independent verification.
Moonshot AI released Kimi K3.1 Pro through its apps, Kimi Work, Kimi Code and API on October 23, and an independent Artificial Analysis evaluation placed it fourth on its Intelligence Index with a score of 55.1. That puts the model three points behind the index leader, but its weights are not yet available: Moonshot has promised a release by November 4, and has not published a licence for them.
Artificial Analysis’s October 24 model-page figures put Claude Opus 5.5 at 57.6, Claude Sonnet 5.5 at 56.0 and Muse Wonder Full at 55.8, ahead of Kimi K3.1 Pro. The results compare models at their highest listed reasoning settings. Kimi’s score is a substantial rise from the earlier Kimi K3 score of 43.6 in the same table, although scores depend on the evaluation and settings used.
Artificial Analysis measured Kimi K3.1 Pro at $2.11 per Intelligence Index task, compared with $5.98 for Opus 5.5 at maximum effort. The listed API prices are $2 per million input tokens and $10 per million output tokens. These figures make the model’s benchmark performance relevant to organisations weighing capability against usage costs, but they do not establish that it will be cheaper for every workload.
The benchmark shows a particularly large gain on agentic coding tasks: Kimi scored 67% on Terminal-Bench 4.0, versus 13% for Kimi K3 in Artificial Analysis’s evaluation. It also scored 1,846 Elo on GDPval-AA, behind Opus 5.5 at 1,866 and Muse Wonder Full at 1,857. Other results were less dominant: Kimi reached 52% on Humanity’s Last Exam, compared with 61% for Opus 5.5. On AA-Omniscience, its measured hallucination rate was 44%, down from 53% for Kimi K3.
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.
Moonshot reports
vendor claims own harness, partly internal tests
| Benchmark | K3.1 Pro | Comparison |
|---|---|---|
| Terminal-Bench 4.0 | 65.7% | vs. Claude Opus 5.5: 66.4% |
| BenchCAD with tools | 92.7% | vs. Claude Opus 5.5: 96.2% |
| OfficeQA Pro v2 | 64.4 | vs. Claude Opus 5.5: 62.2 |
| Kimi Hallu Bench Hard internal | 8.3% | vs. Kimi K3: 23.4% (about −65%) |
Artificial Analysis measured
independent highest reasoning setting per model
| Measurement | K3.1 Pro | Comparison |
|---|---|---|
| Terminal-Bench 4.0 | 67% | vs. Claude Opus 5.5 (max): 60% |
| GDPval-AA | 1,846 Elo | vs. Claude Opus 5.5: 1,866 |
| Humanity’s Last Exam | 52% | vs. Claude Opus 5.5: 61% |
| AA-Omniscience hallucination rate | 44% | 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 Narrower Gap on Agentic Tasks
The result matters because open-weight models have generally scored below leading closed systems in the comparison cited by Artificial Analysis. Its previous leading open-weight entry, Xiaomi’s MiMo-V2.6-Pro, scored 46.3, while Kimi K3.1 Pro reached 55.1. If Moonshot publishes usable weights, the result could give developers and companies a stronger option for adapting or running a model outside a vendor’s hosted service.
That possibility is conditional, not yet a present feature. Users can access Kimi K3.1 Pro through Moonshot’s products and API, but cannot currently download the promised weights. Even after release, the model’s reported 2.8 trillion total parameters and 104 billion active parameters mean self-hosting would require substantial computing resources. The 55.1 score also does not show that the model is best for every task: its clearest advantage in the reported figures is agentic work, while knowledge and other benchmark results vary.
As an affiliate, we earn on qualifying purchases.
What the Benchmarks Actually Measure
Artificial Analysis describes Kimi K3.1 Pro as a more heavily post-trained version of Kimi K3. Its listing gives it a one-million-token context window and text and image input. The Intelligence Index is a composite score, so its ranking offers a comparative snapshot across evaluations rather than a guarantee of performance in a particular company’s software, data or workflow.
Moonshot’s launch material makes broader claims, including that the model can complete projects in Unity, Unreal Engine and Godot, create parametric 3D models, and perform finance work. Those demonstrations and many of the supporting results are vendor-reported, not independent confirmation of reliable everyday use. Moonshot, for example, reports a 92.7% result on BenchCAD with tools, below the reported 96.2% for Opus 5.5 and 95.9% for GPT-6 Astra.
The claims about hallucinations also use different tests. Moonshot says its internal Kimi Hallu Bench Hard recorded claim-level hallucinations falling from 23.4% for Kimi K3 to 8.3% for K3.1 Pro, a roughly 65% reduction in that test. It describes the benchmark as deliberately designed to provoke errors and says the results do not represent overall daily-use error rates. Artificial Analysis measured a smaller fall on its own AA-Omniscience evaluation, from 53% to 44%.
“The internal hallucination test is “deliberately crafted” to provoke errors, and its rates “do not represent overall error rates in daily use.””
— Moonshot AI
GPU workstation for machine learning development
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Weights and Licence Still Pending
The central open question is whether Moonshot will publish the weights by its stated November 4 target, and under what licence. Without the files and terms, developers cannot confirm whether the release will permit commercial use, modification or redistribution. Artificial Analysis’s description of Kimi as a future open-weight leader depends on that release taking place.
Benchmark scores also leave practical questions unresolved. The supplied results do not establish how the model performs across a broad set of independent, real-world deployments, or how much its results vary by prompt, tools and task. Moonshot’s product demonstrations and internal tests should not be treated as equivalent to third-party validation. It is also not clear what hardware and operating costs would be needed for organisations seeking to run a model of this scale themselves.
As an affiliate, we earn on qualifying purchases.
Watch for the November Weights Release
The next concrete milestone is Moonshot’s promised November 4 release of Kimi K3.1 Pro’s weights. Developers and companies will be able to assess the files, licence terms and practical hardware requirements only if the release arrives. Further independent testing could then show whether the benchmark performance holds outside hosted access and Moonshot’s own demonstrations.
Until then, Kimi K3.1 Pro is available through Moonshot’s listed services, and the reported score and prices offer a basis for comparison—not proof of a general performance or cost advantage. Readers should look for the actual licence and independent evaluations before treating the model as an open-weight alternative they can deploy on their own systems.
cloud computing and machine learning reference books
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Kimi K3.1 Pro’s Artificial Analysis score?
Artificial Analysis listed a 55.1 Intelligence Index score on October 24, 2026, ranking it fourth in the comparison provided. Its reported score was below Opus 5.5 at 57.6, Sonnet 5.5 at 56.0 and Muse Wonder Full at 55.8.
Can users download and run Kimi K3.1 Pro now?
No. Moonshot made the model available through its apps, Kimi Work, Kimi Code and API, but said its weights would be released by November 4. The licence had not been published in the source material.
What does “frontier-near” mean in this report?
It describes Kimi’s proximity to the leading models in the cited Artificial Analysis comparison: its score of 55.1 was three points below the 58-point rounded score for the leader in the source’s summary. It does not mean the model leads every benchmark or matches every proprietary system on all tasks.
Are Moonshot’s claims about fewer hallucinations independently verified?
Not as stated. Moonshot reported a roughly 65% reduction on its internal, deliberately challenging test. Artificial Analysis reported a smaller change on its separate evaluation, with the measured rate falling from 53% to 44%.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
