# Kimi K3.1 Pro Pricing, Evaluation Scores, and Open Weights

URL: https://3ammarketer.com/news/kimi-k3-1-pro-open-weights-pricing-benchmarks
Published: 2026-10-30

> Moonshot announced Kimi K3.1 Pro with reduced API pricing, an 8.3% hallucination rate on internal tests, and open model weights arriving November 4, 2026.

Moonshot announced Kimi K3.1 Pro on October 23, 2026, cutting API prices to $2.00 per million input tokens and $10 per million output tokens while scheduling full open weights for release by November 4, 2026. [1][2]

Moonshot reported that claim-level hallucination on its internal Kimi Hallu Bench Hard stress test dropped from 23.4% on K3 down to 8.3% on K3.1 Pro. [1]

Hallucination rate on the Kimi Hallu Bench Hard stress test (%). [1]

- K3: 23.4 %
- K3.1 Pro: 8.3 %

## Performance and published evaluations

As of October 23, 2026, the Artificial Analysis Intelligence Index ranked Kimi K3.1 Pro fourth with a score of 55. Moonshot also made Kimi K3.1 Pro Highspeed available, which delivers 2 to 3 times the speed of the standard K3.1 Pro model. [3][1]

## The cost of safety in model training

To reduce misuse risks ahead of public weight distribution, Moonshot identified and removed cyber-related training data from both the mid-training and reinforcement learning stages. [1]

On the internal xdev-bench-hard benchmark, an unfiltered checkpoint without data removal or RL filtering solved 90.5% of problems at pass@3. The safety-filtered K3.1 Pro dropped to 81.0%, showing an explicit trade-off between open distribution safety and high-end software development execution. [1]

Pass@3 performance on the xdev-bench-hard benchmark (%). [1]

- Unfiltered: 90.5 %
- Safety-filtered: 81 %

## Pricing shifts from K3 to K3.1 Pro

Moonshot lowered API costs across both input and output tiers. According to data tracked on our [open model release tracker](/data/open-models), the original Kimi-K3 launched on June 13, 2026, with 2779.9B parameters under the kimi-k3 licence. [4]

Our pricing logs captured on October 13, 2026, recorded kimi-k3 at $3 input and $15 output per million tokens. The shift to $2.00 input and $10 output is a cut of a third on both. [5][2]

API pricing comparison for input and output tokens ($ per 1M tokens). [1][5]

- K3 Input: 3 $
- K3.1 Pro Input: 2 $
- K3 Output: 15 $
- K3.1 Pro Output: 10 $

Our read: If your workflows prioritize factual accuracy and general agentic workflows, switch to K3.1 Pro immediately at $2.00 per million input tokens. Keep specialized coding agents on alternative models if you cannot accept the drop from 90.5% to 81.0% on tough development tasks.

Limit: The reported hallucination reductions and development benchmark regressions rely on Moonshot's internal tests, specifically Kimi Hallu Bench Hard and xdev-bench-hard. Independent multi-tenant evaluations on production engineering workloads have not yet been verified.

## Sources

1. [Kimi K3.1 Pro Tech Blog: Advancing Open Intelligence](https://www.kimi.ai/blog/kimi-k3-1-pro)
2. [Introducing Kimi K3.1 Pro](https://simonwillison.net/2026/Oct/23/kimi-k31-pro/)
3. [Kimi K3.1 Pro launches at 4th on the Intelligence Index and will lead open weights models | Artificial Analysis](https://artificialanalysis.ai/articles/kimi-k3-1-pro-launch-benchmarks)
4. [Open model release tracker (3AM Marketer data set)](https://3ammarketer.com/data/open-models)
5. [LLM API pricing timeline (3AM Marketer data set)](https://3ammarketer.com/data/model-pricing)