# We shrank Kimi K3 by 35%, within a point of the original on MMLU

Tileward research report R-011. Kimi K3 is Moonshot AI's 2.8-trillion-parameter model. As published its weights are 1,560.9 GB, already stored in 4 bits, and it ran here on eight B300 cards. Tileward's TWC25 build is 1,007.9 GB, 35% smaller (1.5×), and TWMX34 is 1,178.1 GB, 25% smaller (1.3×). On MMLU TWC25 is 0.6 points below the original, inside the one-point bar set before the run, and TWMX34 is level. Both are level on made-up facts and within a few calls of the original on tool calling.

Page: https://tileward.com/research/a-smaller-kimi-k3/

## How it was measured

Every build ran in one session, on the same machine and engine, and answered the same questions.

- **Builds:** [Kimi K3 as published](https://huggingface.co/moonshotai/Kimi-K3) and two Tileward builds: TWMX34 and TWC25. The published model is already 4-bit, so the percentages are against a 4-bit model, not a 16-bit one.
- **Machine:** eight B300 cards, with stock vLLM serving each build from its own weights, split across all eight, with a 16-bit cache and 90% of each card.
- **Size:** the weights files, as the model library lists them.
- **MMLU:** [MMLU](https://arxiv.org/abs/2009.03300), five-shot, no sampling, scored by the answer letter the model rates most likely. The bar, set before the run: a build passes if the paired test puts it no more than one point below the original. The paired test counts the answers that differ: 290 right only for the original and 207 only for TWC25; 200 only for the original and 185 only for TWMX34.
- **Made-up facts:** each build reads 360 passages from Tileward's own document set and writes out the facts they state. Every fact is matched back to its passage, and one that no passage supports counts as made up. The test treats facts as independent.
- **Tool calls:** six cases, 20 trials each, run five times: 600 calls, once at low thinking effort and once at high. Each call is scored on tool name, call count and argument shape, with vLLM's Kimi K3 tool parser.
- **Agreement:** each build reads the same 300 passages (82,712 token positions), and we count where its top next token is the original's.

## Results

| Build | Weights | Against the original |
|---|---|---|
| Kimi K3 as published | 1,560.9 GB | The reference |
| Tileward TWMX34 | 1,178.1 GB | 25% smaller (1.3×) |
| Tileward TWC25 | 1,007.9 GB | 35% smaller (1.5×) |

| Test | The original | Tileward TWMX34 | Tileward TWC25 |
|---|---|---|---|
| MMLU, 14,042 questions | 91.1% | 91.0% | 90.5% |
| MMLU against the original | – | Level (p = 0.48) | 0.6 points below (p = 0.0002), inside the bar |
| Facts made up | 1.0% (13 of 1,350) | 0.6% (8 of 1,323) | 0.9% (12 of 1,384) |
| Tool calls right, low thinking effort | 598 of 600 | 593 of 600 | 596 of 600 |
| Tool calls right, high thinking effort | 597 of 600 | 590 of 600 | 595 of 600 |

Neither build differs from the original on made-up facts by more than chance (p = 0.38 and 0.84). On MMLU the paired test puts TWC25 at most 0.85 points below the original, and TWMX34 at most 0.34.

Next-word agreement with the original:

| Build | Next-word agreement | KL distance from the original |
|---|---|---|
| Tileward TWMX34 | 90.8% | 0.036 |
| Tileward TWC25 | 88.6% | 0.060 |

Memory, the server allowed 90% of each card:

| Build | Weights on each card | Tokens the cache holds |
|---|---|---|
| Kimi K3 as published | 191.7 GiB | 1,467,558 |
| Tileward TWMX34 | 147.1 GiB | 3,139,006 |
| Tileward TWC25 | 127.3 GiB | 3,945,603 |

## Which build

TWMX34 is level with the original on MMLU and agrees with it more often. TWC25 gives up 0.6 points of MMLU for 170.2 GB less weight than TWMX34 and 26% more room in the cache. A third build at TWC25's size missed the one-point bar on MMLU and fell further behind on tool calls, so it is not offered.

## Related

- [Kimi K3 eval card](https://tileward.com/evals/#compressed-model-k3): the result in one screen
- [Kimi K3 on Hugging Face](https://huggingface.co/moonshotai/Kimi-K3): the published model
- [Talk to an engineer](https://calendar.app.google/hg2DzFTxKJgwFWm19): bring the work you would run on Kimi K3 and the cards you run on
- [Research](https://tileward.com/research/): every report