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).

AI & Machine Learning

Explore on GitHub Read Hacker News discussions

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 7–Nov 14 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/14/2026, 5:04:57 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.

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.
Period	Start (UTC)	End (exclusive, UTC)	Mentions	Comments
Jul 19, 00:00	2026-07-19 00:00:00 UTC	2026-07-19 06:00:00 UTC	2	2
Jul 19, 06:00	2026-07-19 06:00:00 UTC	2026-07-19 12:00:00 UTC	0	0
Jul 19, 12:00	2026-07-19 12:00:00 UTC	2026-07-19 18:00:00 UTC	0	0
Jul 19, 18:00	2026-07-19 18:00:00 UTC	2026-07-20 00:00:00 UTC	0	0
Jul 20, 00:00	2026-07-20 00:00:00 UTC	2026-07-20 06:00:00 UTC	0	0
Jul 20, 06:00	2026-07-20 06:00:00 UTC	2026-07-20 12:00:00 UTC	1	1
Jul 20, 12:00	2026-07-20 12:00:00 UTC	2026-07-20 18:00:00 UTC	0	0
Jul 20, 18:00	2026-07-20 18:00:00 UTC	2026-07-21 00:00:00 UTC	0	0
Jul 21, 00:00	2026-07-21 00:00:00 UTC	2026-07-21 06:00:00 UTC	0	0
Jul 21, 06:00	2026-07-21 06:00:00 UTC	2026-07-21 12:00:00 UTC	0	0
Jul 21, 12:00	2026-07-21 12:00:00 UTC	2026-07-21 18:00:00 UTC	0	0
Jul 21, 18:00	2026-07-21 18:00:00 UTC	2026-07-22 00:00:00 UTC	0	0
Jul 22, 00:00	2026-07-22 00:00:00 UTC	2026-07-22 06:00:00 UTC	0	0
Jul 22, 06:00	2026-07-22 06:00:00 UTC	2026-07-22 12:00:00 UTC	1	1
Jul 22, 12:00	2026-07-22 12:00:00 UTC	2026-07-22 18:00:00 UTC	0	0
Jul 22, 18:00	2026-07-22 18:00:00 UTC	2026-07-23 00:00:00 UTC	0	0
Jul 23, 00:00	2026-07-23 00:00:00 UTC	2026-07-23 06:00:00 UTC	0	0
Jul 23, 06:00	2026-07-23 06:00:00 UTC	2026-07-23 12:00:00 UTC	0	0
Jul 23, 12:00	2026-07-23 12:00:00 UTC	2026-07-23 18:00:00 UTC	1	1

What Hacker News says

5 sources

chlee·Jul 23, 2026

the rumor in chinese web is the following. Z.ai found a way to crack/overcome the cryptography that safeguarded Claude/Fable's thinking tokens in Claude Code, so Z.ai had the ability to peak into decrypted thinking tokens that Fable produced in Claude Code. Z.ai also shared their findings with other big LLM companies like Moonshot AI, which had 20,000+ accounts speed running through Fable to get their hands on as many high quality thinking tokens as possible for distillation. https://github.com/rgreenblatt/which_claude_is_k3/blob/main/... https://x.com/alex_prompter/status/2079625670918046048 Like all rumors, who knows what's true or what's not.

In OpenAI and Anthropic unite against open-weight AI risks to their bottom line

tristanj·Jul 22, 2026

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

tristanj·Jul 20, 2026

here's a chart of asking various models to self-identify: https://github.com/rgreenblatt/which_claude_is_k3/blob/main/... Qwen is remarkably consistent, correctly identifying itself 100% of the time. Kimi K3 performs poorly at this test, it correctly self-identifies ~80% of the time; sometimes calling itself Claude and rarely ChatGPT.

In The Kimi K3 Moment

tristanj·Jul 19, 2026

API distillation doesn't have to explain all of K3's capabilities for it to have happened. Kimi K3 reproducibly identifies itself as Claude: https://x.com/denisewu/status/2077984660211269870 This behavior is exactly what you'd expect from a model distilled from Claude. There's a detailed analysis of K3's ambiguous identity here: https://github.com/rgreenblatt/which_claude_is_k3/blob/main/... This analysis observed K3 identifies itself as Claude approximately 15% of the time. K3 reproduces Claude's correct current model id, which the real Claude models themselves do not emit. This suggests K3 was trained on Claude data labeled with deployment metadata (API logs, tagged synthetic data), rather than Claude's chat outputs. And there's an entire Reddit thread discussing Kimi's similarities with Claude https://www.reddit.com/r/LocalLLaMA/commen

Read the rest

ts/1m2w5ge/did_kim... This analysis shows K3 and Opus/Fable have unexpected correlated outputs https://typebulb.com/u/lab/you-re-relatively-right/full

In The Kimi K3 Moment

tristanj·Jul 19, 2026

There's reproducible evidence of Kimi K3 spontaneously identifying itself as Claude https://x.com/denisewu/status/2077984660211269870 This behavior is exactly what you'd expect from a model distilled from Claude. Someone even took the time to analyze Kimi's ambiguous identity, in great detail: https://github.com/rgreenblatt/which_claude_is_k3/blob/main/... And there's an entire Reddit thread discussing this https://www.reddit.com/r/LocalLLaMA/comments/1m2w5ge/did_kim... That doesn’t prove Anthropic’s specific 3.4m-session allegation, but calling it “zero evidence” is no longer credible. Kimi K2.5 was worse in a hilarious way, it identified itself as Claude and referenced Anthropic's Constitutional AI as some of its guiding principles https://huggingface.co/moonshotai/Kimi-K2.5/discussions/38

In The Kimi K3 Moment