# Moonshot publishes Kimi K3 weights: 2.8T parameters

> Moonshot published Kimi K3's 2.8-trillion-parameter weights on 27 July 2026.

*The largest open-weight release ever, eleven days after the hosted launch.*

By The FeaturedDaily Desk · FeaturedDaily
Canonical: https://featureddaily.com/news/kimi-k3-weights-released-2-8-trillion

Moonshot AI published the complete weights for Kimi K3 on 27 July 2026, eleven days after launching it as a hosted service. The Hugging Face repository holds 96 shards, the licence and the technical report.

K3 is a Mixture-of-Experts model with 2.8 trillion total parameters and 104 billion active. Sixteen of 896 routed experts fire per token alongside two shared experts. Context is 1,048,576 tokens. Quantisation-aware training runs from the supervised fine-tuning stage using MXFP4 weights and MXFP8 activations.

## Benchmark position

| Benchmark | K3 | Comparison |
| --- | --- | --- |
| Arena Frontend Code | 1,679 (1st) | Ahead of Claude Fable 5 |
| AA-Briefcase | 1,527 (2nd) | GPT-5.6 Sol Max 1,495 |
| GDPval-AA v2 | 1,687 (3rd) | Claude Fable 5 Max 1,815 |
| FrontierSWE | 81.2 | Claude Fable 5 86.6 |

> On AA-Briefcase: second place with a score of 1,527 — beating GPT-5.6 Sol Max.
> — [VentureBeat](https://venturebeat.com/technology/chinas-moonshot-ai-releases-kimi-k3-the-largest-open-source-model-ever-rivaling-top-u-s-systems), 2026-07-27

## Deployment and licence

Moonshot recommends at least 64 accelerators. A four-bit estimate puts weights alone near 1.4TB. The licence is a custom Kimi K3 License; reports describing it as MIT are inaccurate.

Pricing is $3.00 per million cache-miss input tokens, $0.30 cache-hit, and $15.00 per million output, flat across the context window. Claude Fable 5.1 lists at $10 and $50.

The weekend after the release, Alibaba announced a 2.4-trillion-parameter Qwen 3.8 with open weights, a class it had previously kept API-only. Anthropic's February distillation allegation against Moonshot remains unresolved.

## Known limitations

Moonshot acknowledges three constraints. The model is sensitive to thinking-history truncation, so harnesses that trim chain-of-thought degrade output quality. It shows excessive proactiveness in ambiguous scenarios. And conversational polish trails Claude Fable 5 and GPT-5.6 Sol.

K3 launched with a single reasoning effort level running at maximum thinking by default; low and high modes were promised in later updates. Most detailed scores are vendor-reported using mixed agent harnesses, so public weights improve inspectability without normalising the tests.

K3 ranked first on four of eight real-world task automation benchmarks, including Automation Bench, SpreadsheetBench 2 and BrowseComp, where it posted a state-of-the-art 91.2 out of 100.

## Key takeaways

- Moonshot published the full Kimi K3 weights on 27 July 2026, eleven days after the 16 July hosted launch.
- 2.8 trillion total parameters, 104 billion active, 1,048,576-token context — the largest open release to date.
- First on Arena's Frontend Code evaluation at 1,679, ahead of Claude Fable 5, in blind developer testing.
- Ships under a custom Kimi K3 License, not MIT. Moonshot recommends 64+ accelerators to serve it.
- API pricing is $3 per million cache-miss input tokens and $15 per million output.

## FAQ

### How big is Kimi K3?
2.8 trillion total parameters with 104 billion active — the largest open-weight model released to date.

### What licence does it use?
A custom Kimi K3 License, not MIT.

### What hardware does it need?
Moonshot recommends at least 64 accelerators; weights alone are roughly 1.4TB at four-bit precision.

## Sources

- [China's Moonshot AI releases Kimi K3, the largest open-source model ever, rivaling top U.S. systems](https://venturebeat.com/technology/chinas-moonshot-ai-releases-kimi-k3-the-largest-open-source-model-ever-rivaling-top-u-s-systems) — VentureBeat, 2026-07-27
- [Moonshot AI releases weights for Kimi-K3, firing a shot across the bow of OpenAI and Anthropic](https://www.tomshardware.com/tech-industry/artificial-intelligence/moonshot-ai-releases-weights-for-kimi-k3-firing-a-shot-across-the-bow-of-openai-and-anthropic-open-weight-model-performs-almost-as-well-as-frontier-models-while-being-2-3x-easier-to-run) — Tom's Hardware, 2026-07-27
- [China's 2.8-trillion-parameter Kimi K3 beats Claude Fable 5 in Frontend Code Arena benchmark](https://www.tomshardware.com/tech-industry/artificial-intelligence/moonshot-releases-2-8-trillion-parameter-kimi-k3) — Tom's Hardware, 2026-07-17
- [moonshotai/Kimi-K3 — model card, benchmark tables and Kimi K3 License](https://huggingface.co/moonshotai/Kimi-K3) — Moonshot AI (Hugging Face), 2026-07-27
- [Kimi K3: The open-weights escalation](https://www.interconnects.ai/p/kimi-k3-the-open-weights-escalation) — Interconnects, 2026-07-20