MoonshotAI: Kimi K3
moonshotai/kimi-k3 · Moonshot AINot yet established- Declared family
- Unknown
- Evidence confidence
- Unknown
- Declared tokenizer tag
- Other
- Entered catalogue
- 16 July 2026
- Last tested
- 21 August 2026
- Measurement
- Measured
Identity statement
MoonshotAI: Kimi K3 has a clean measured fingerprint that currently sits alone: no other measured model shares its signature.
Evidence stack
Signals of different kinds, weighed together| Signal | Result | Reference |
|---|---|---|
| Tokenizer signature | Measured with 1 corrupt row excluded | – |
| API surface | No other entry shares this exact contract | api_7600defa |
| Context and output | 1,048,576 context · – max output | declared |
| Reasoning contract | Optional · efforts max, high, low · default max | declared |
| Serving providers observed | DeepInfra, Moonshot AI, Chutes, Modal, Together, DigitalOcean, Fireworks | measured 21 August 2026 |
Closest measured models
Agreement across the mutually clean test stringsTencent: Hy-MT2-1.8Btencent/hy-mt2-1.8b30/49 · 61%Tencent: Hy-MT2-30B-A3Btencent/hy-mt2-30b-a3b30/49 · 61%Tencent: Hy3tencent/hy330/49 · 61%Tencent: Hy3 previewtencent/hy3-preview30/49 · 61%DeepSeek: DeepSeek V4 Pro 0813deepseek/deepseek-v4-pro-081325/49 · 51%DeepSeek: DeepSeek V4 Flash 0731deepseek/deepseek-v4-flash-073125/49 · 51%
Click a row to open the full comparison.
The measured fingerprint
Marginal prompt-token cost of each test string, grouped by scriptAll 50 rows, 1 excluded as corrupt
English and whitespace
en-prose14
en-long16
spaces-203
spaces-603
tabs-204
newlines-204
mixed-ws6
Digits
digits-93
digits-124
digits-3010
digits-sep7
float-long9
CJK
zh-common9
zh-long10
zh-rare18
ja-kana11
ja-kanji14
ko13
Other scripts
ru16
ar15
he16
hi14
th18
el57
Emoji
emoji-basic7
emoji-skin20
emoji-zwj-family11
emoji-zwj-x333
emoji-flags14
emoji-prof20
Rare Unicode
math26
boxdraw24
combiningx
cjk-ext-b13
surrogates31
zalgo36
rtl-mix8
Code
py-code25
py-indent15
json20
html16
regex40
camel6
snake3
Repetition and encodings
rare-word-x528
repeat-tok41
base6430
hex11
url17
uuid27
An x marks a row excluded as corrupt (caching or provider interference detected during measurement). Overhead subtracted: 88 prompt tokens.
Declared record
What the catalogue claims about this modelKimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at...
- Modalities
- text+image+video->text
- Declared tokenizer
- Other
- Prompt price
- $3.00 / M tokens
- Completion price
- $15.0 / M tokens
Supported parameters
frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_ppresence_penaltyreasoningreasoning_effortrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p
defaults: {"temperature":null,"top_p":0.95,"top_k":null,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":null}
History
Every observation, kept as taken- 16 July 2026Enters the OpenRouter catalogue with no declared family.
- 21 August 2026Fingerprinted in the YFarmX catalogue sweep · 49 of 50 strings measured clean.
Cite this page as the evidence record for MoonshotAI: Kimi K3: the URL is stable, measurements are dated, and revisions append to the history above. Method: tokenizer fingerprinting, v1, 50 strings. The raw data behind every figure is on the data page.