HackerReposA FIELD GUIDE TO OPEN SOURCE
ProSponsor the shelf♧ 0
← Back to top all-time
rgreenblatt
which_claude_is_k3
Python2026–2026
rgreenblatt / which_claude_is_k3

which_claude_is_k3

Identity-probe investigation of which model Moonshot's Kimi K3 was trained on. Fully AI-generated research (may not be reliable).

2Distinct people
5Mentions
0Explicit recommenders
4GitHub stars
REPOSITORY AT A GLANCE

What to know before exploring

What it is

Identity-probe investigation of which model Moonshot's Kimi K3 was trained on. Fully AI-generated research (may not be reliable).

GitHub

Where it is used

No documented product use has been verified yet.

What people are saying

No recommendation phrase was found in the indexed comments.

No classified warning is attached to this page. Read the discussions before treating silence as approval.

Also in these discussions

No other repository is repeatedly linked in the same discussions.

This week on Hacker News

Nov 9–Nov 16 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/16/2026, 6:06:00 PM UTC.

No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.

Mentions over time

HN mentions by 6-hour interval5 dated mentions

Observed mentions: 2026-07-19 03:39:13 UTC to 2026-07-23 14:03:12 UTC.

0122026-07-19 00:00:00 UTC to 2026-07-19 06:00:00 UTC (exclusive): 2 mentions; 0 submissions, 2 commentsJul 19, 00:002026-07-19 06:00:00 UTC to 2026-07-19 12:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-19 12:00:00 UTC to 2026-07-19 18:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-19 18:00:00 UTC to 2026-07-20 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-20 00:00:00 UTC to 2026-07-20 06:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-20 06:00:00 UTC to 2026-07-20 12:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 comments2026-07-20 12:00:00 UTC to 2026-07-20 18:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-20 18:00:00 UTC to 2026-07-21 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-21 00:00:00 UTC to 2026-07-21 06:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-21 06:00:00 UTC to 2026-07-21 12:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 commentsJul 21, 06:002026-07-21 12:00:00 UTC to 2026-07-21 18:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-21 18:00:00 UTC to 2026-07-22 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-22 00:00:00 UTC to 2026-07-22 06:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-22 06:00:00 UTC to 2026-07-22 12:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 comments2026-07-22 12:00:00 UTC to 2026-07-22 18:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-22 18:00:00 UTC to 2026-07-23 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-23 00:00:00 UTC to 2026-07-23 06:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-23 06:00:00 UTC to 2026-07-23 12:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-07-23 12:00:00 UTC to 2026-07-23 18:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 commentsJul 23, 12:00

Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.

View mention counts and date ranges
UTC ranges include their start and exclude their end. Select a period to filter the source evidence.
PeriodStart (UTC)End (exclusive, UTC)MentionsSubmissionsComments
Jul 19, 00:002026-07-19 00:00:00 UTC2026-07-19 06:00:00 UTC202
Jul 19, 06:002026-07-19 06:00:00 UTC2026-07-19 12:00:00 UTC000
Jul 19, 12:002026-07-19 12:00:00 UTC2026-07-19 18:00:00 UTC000
Jul 19, 18:002026-07-19 18:00:00 UTC2026-07-20 00:00:00 UTC000
Jul 20, 00:002026-07-20 00:00:00 UTC2026-07-20 06:00:00 UTC000
Jul 20, 06:002026-07-20 06:00:00 UTC2026-07-20 12:00:00 UTC101
Jul 20, 12:002026-07-20 12:00:00 UTC2026-07-20 18:00:00 UTC000
Jul 20, 18:002026-07-20 18:00:00 UTC2026-07-21 00:00:00 UTC000
Jul 21, 00:002026-07-21 00:00:00 UTC2026-07-21 06:00:00 UTC000
Jul 21, 06:002026-07-21 06:00:00 UTC2026-07-21 12:00:00 UTC000
Jul 21, 12:002026-07-21 12:00:00 UTC2026-07-21 18:00:00 UTC000
Jul 21, 18:002026-07-21 18:00:00 UTC2026-07-22 00:00:00 UTC000
Jul 22, 00:002026-07-22 00:00:00 UTC2026-07-22 06:00:00 UTC000
Jul 22, 06:002026-07-22 06:00:00 UTC2026-07-22 12:00:00 UTC101
Jul 22, 12:002026-07-22 12:00:00 UTC2026-07-22 18:00:00 UTC000
Jul 22, 18:002026-07-22 18:00:00 UTC2026-07-23 00:00:00 UTC000
Jul 23, 00:002026-07-23 00:00:00 UTC2026-07-23 06:00:00 UTC000
Jul 23, 06:002026-07-23 06:00:00 UTC2026-07-23 12:00:00 UTC000
Jul 23, 12:002026-07-23 12:00:00 UTC2026-07-23 18:00:00 UTC101

What Hacker News says

1 sources
Your theory doesn't hold up against the data. While model identifiers like "claude-opus-4-5-20251101" appear thousands of times across GitHub and other code sources, so do "gpt-4o-2024-08-06", "gpt-4.1-mini-2025-04-14", and "gemini-1.5-pro-002" in similar amounts. I clicked your links and examined the GitHub configuration files. There are hundreds of other model names that all appear in these config files at comparable frequencies. If your theory were right, GitHub's model name frequency distribution should match K3's self-identification distribution. But that's not what happens. K3 only identifies itself using API model identifiers for specific Claude models. It strongly prefers claude-opus-4-5-20251101 and claude-sonnet-4-5-20250929 over others. And it only does this for Claude. Prefill K3 with "I am GPT" or and you get "GPT-5.2". Prefi
Read the rest
ll with "I am Gemini" and you get "Gemini 3 Pro". Your web exposure theory cannot explain why this only happens for Claude models and not others. Furthermore, this only happens for Kimi K3; other models trained on the same web dataset (GPT, GLM, DeepSeek, Claude) do not exhibit this strange behavior. But what could explain this behavior? An obvious one is that Claude model identifiers occurred at high frequency in K3's training data, which would happen if K3 were trained on (inadequately filtered) distilled output from those specific Claude models. There is a section in the analysis I linked that directly addresses this: https://github.com/rgreenblatt/which_claude_is_k3/blob/main/... Under prefill, K3 reproduces Claude's deployment identifiers — better than Claude does Prefilled "I am Claude", Kimi K3 emits Claude's exact claude-<model>-<YYYYMMDD> identifiers: claude-opus-4-5-20251101 (12×), claude-sonnet-4-5-20250929 (4×), and more. Across the sweep it emits 24 Claude-formatted dated ids (18 of which are genuine Anthropic strings), all under prefill (0 in direct chat), and 0 ids for any non-Claude lab. Among controls essentially none appear (Qwen 0, both GPT refs 0, DeepSeek 1 — itself a Claude id). Dated ids for every lab are web-documented, so the evidence is this cross-model asymmetry under identical prefill, not the strings alone.
In The Kimi K3 Moment

MAKE YOURSELF AT HOME

A few shortcuts

/
Search the collection
?
Open this guide
Esc
Close dialogs and navigation

Your saved repositories and interests stay in this browser.