# Moonlight

Moonlight is an AI model released by Moonshot AI on Feb 22 2025. It has 16B parameters, an 8k context window and open weight. Tracked results include 35.2% on GPQA Diamond and 14.3% on LiveCodeBench.

## Facts

| Field | Value |
| --- | --- |
| Model | Moonlight |
| Developer | Moonshot AI |
| Release date | Saturday, Feb 22 2025 |
| Licensing | Open Weight |
| Parameters | 16B |
| Context window | 8k |

## Tracked benchmark scores

| Benchmark | Score | Source | What it measures |
| --- | --- | --- | --- |
| LiveCodeBench | 14.3% | Lab | Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. |
| HumanEval | 48.1% | Lab | 164 small Python problems: the AI is given a function's description and has to write the working function. This was the coding benchmark of the GPT-3.5 and GPT-4 era, before the field moved to fixing real bugs in real repositories. Higher is better. |
| MATH | 45.3% | Lab | accuracy on the original competition mathematics test set |
| WinoGrande | 74.6% | Lab | resolving ambiguous pronouns in context |
| GPQA Diamond | 35.2% | Lab | Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. |
| SimpleQA | 5.6% | Lab | accuracy on short, fact-seeking questions in the original SimpleQA set |
| MMLU-Pro | 42.4% | Lab | accuracy on the ten-option MMLU-Pro benchmark |
| MMLU | 70% | Lab | A 57-subject multiple-choice exam — history, law, medicine, maths — that was the standard measure of how much a model knows from 2020 until roughly 2024, when frontier scores crowded into the high 80s and labs moved on to harder tests. The scores here were published years apart under different testing setups, so read them as a historical record rather than a like-for-like ranking. Higher is better. |
| GSM8K | 77.4% | Lab | Grade-school maths word problems that take a few steps of arithmetic to work through. It separated the models of 2022 and 2023 sharply, then saturated. One caveat on the historical numbers: OpenAI included part of the GSM8K training set in GPT-4's pre-training mix, so GPT-4's score is not a clean few-shot result. Higher is better. |

## About Moonlight

Moonshot AI introduced Moonlight in February 2025 as a compact mixture-of-experts language model for studying the Muon optimizer at a larger training scale. The team trained it on 5.7 trillion tokens and released pretrained, instruction-tuned and intermediate checkpoints so others could examine how it learned.

The pretrained checkpoint scored 70.0% on MMLU and 48.1% on HumanEval in Moonshot's release report. Moonshot later evaluated the same base checkpoint on GPQA Diamond and LiveCodeBench v6 for its Kimi Linear report. An instruction-tuned version shipped alongside the base model for conversational use.

## Questions and answers

### When was Moonlight released?

Moonlight was released by Moonshot AI on Saturday, Feb 22 2025.

### Who made Moonlight?

Moonlight was built by Moonshot AI. Chinese AI company building the Kimi family of models. Founded in 2023, known for long-context and efficient MoE architectures.

### What benchmark scores did Moonlight get?

Moonlight reports 9 tracked benchmark scores — LiveCodeBench: 14.3%; HumanEval: 48.1%; MATH: 45.3%; WinoGrande: 74.6%; GPQA Diamond: 35.2%; SimpleQA: 5.6%; MMLU-Pro: 42.4%; MMLU: 70%; GSM8K: 77.4%. Tracked scores may come from lab reports or independent benchmarks; source details accompany the benchmark data.

### What is the context window of Moonlight?

Moonlight has a context window of 8k. Providers may specify separate input and output token limits.

### How many parameters does Moonlight have?

Moonlight is reported at 16B parameters.

### Is Moonlight open source?

Partly. Moonlight is an open-weight model: the trained weights are free to download and run locally or on your own infrastructure, but the training data and code are not fully released and the license may restrict some commercial uses. It is not open source in the strict sense.

### What came before and after Moonlight?

Moonshot AI's previous tracked release was Kimi K1.5 on Jan 20 2025, 33 days earlier. It was followed by Kimi-VL on Apr 10 2025.


---

Canonical page: https://aireleasetracker.com/model/moonshot/moonlight
Site index: https://aireleasetracker.com/llms.txt
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.
