# Moonshot's Kimi K3 Model Leads in Benchmark

Open-source Kimi K3 outperforms Anthropic's Fable 5.

- **Source:** [Latest news](https://brifing.si/source/77d15218-219f-472d-bf40-31b5646d2942)
- **Published:** 2026-07-17
- **Topic:** [Data & AI](https://brifing.si/c/data-ai)
- **Original article:** <https://www.zdnet.com/article/ai-model-release-tracker/>
- **On Brifing:** https://brifing.si/s/f0be8083-e0b5-4c1f-a8f6-6b3a0acc0585

## Competition Among AI Models

AI labs are continuously developing new models. However, not every new model is a significant leap forward, even if it is better and faster. The strengths of models emerge in context.

## Features of Kimi K3

Developed by Moonshot, Kimi K3 is the largest open-source model on the market with 2.8 trillion parameters. It is designed for long-horizon coding, knowledge work, and reasoning. It is even larger than DeepSeek V4 Pro.

## Benchmark Performance

Kimi K3 outperformed Anthropic's Fable 5 in the Arena benchmark, which measures complex coding tasks. However, it did not surpass Fable 5 in overall performance.

## Open Source Advantage

Kimi K3's competitive performance and open-source nature have reignited the debate on whether American proprietary models are worth the cost. Fable 5 is more expensive than Kimi K3.

## Security Concerns

Open-source models lack the safety guardrails that proprietary models have. These risks are amplified by the fact that Moonshot is a Chinese startup.

## Launch of GPT-5.6

OpenAI released the GPT-5.6 family, which includes Sol, Terra, and Luna models. Sol outperformed Fable 5 in UC Berkeley's Agents Last Exam benchmark.

## Government's Role

The government's involvement in the release of GPT-5.6 indicates a partnership between labs and CAIS. This began with the voluntary testing agreement under the Trump administration.

## Innovations in Muse Spark 1.1

Meta launched Muse Spark 1.1, which outperformed Anthropic's Opus 4.8 and OpenAI's GPT-5.5 in job tasks. It is particularly relevant for white-collar jobs.

## Personal Superintelligence

Muse Spark 1.1 focuses on Mark Zuckerberg's vision of 'personal superintelligence.' This model is designed for personal use rather than professional applications.

## Innovations in GPT-Live-1

OpenAI launched GPT-Live-1, which enables ChatGPT Voice to offer a more realistic conversation experience. Users can seamlessly interrupt the assistant.

## Ease of Access

GPT-Live-1 and GPT-Live-1 mini are available on ChatGPT for Android, iOS, and browsers. This provides significant convenience for users.

## Future of the AI Race

The continuous development of new models is bringing the AI race to an interesting point. This could be a significant turning point in terms of both performance and cost.

## Sources covering this story

- https://www.zdnet.com/article/ai-model-release-tracker/ — 

---

_Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._