llama3-8b-spaetzle-v39

llama3-8b-spaetzle-v39 is a merge of the following models:

    - cstr/llama3-8b-spaetzle-v38
    - cstr/llama3-8b-spaetzle-v36
    - cstr/llama3-8b-spaetzle-v33
    - cstr/llama3-8b-spaetzle-v28

Evaluation

Quantised as q4km gguf, it achieves

on EQ Bench (v2_de): 65.12 (171 parseable),
on EQ Bench (v2):    73.64 (171 parseable).

Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
llama3-8b-spaetzle-v39	43.43	73.31	60	45.89	55.66

GPT4All

Task	Version	Metric	Value		Stderr
arc_challenge	0	acc	60.75	±	1.43
		acc_norm	63.40	±	1.41
arc_easy	0	acc	85.10	±	0.73
		acc_norm	83.46	±	0.76
boolq	1	acc	84.74	±	0.63
hellaswag	0	acc	62.23	±	0.48
		acc_norm	80.73	±	0.39
openbookqa	0	acc	38.00	±	2.17
		acc_norm	46.80	±	2.23
piqa	0	acc	79.92	±	0.93
		acc_norm	80.25	±	0.93
winogrande	0	acc	73.80	±	1.24

Sample output

>>> What is 7777 + 3333?
Let's do the math:

7777 + 3333 = 11110

>>> I have 10 apples. I find 3 gold coins in the bottom of a river. The river runs near a big city that has something to 
... do with what I can spend the coins on. I then lose 4 apples but gain a gold coin. Three birds run into my path and dr
... op 6 apples each. I play an online game and win 6 gold coins but I have to share them equally with my 2 teammates. I 
... buy apples for all the coins I have. The price of an apple is 0.5 coins. How many apples do I have? And where is the 
... river?
Let's break this down step by step!

Initially, you have 10 apples.

You find 3 gold coins in the river.
No mention of spending these coins yet.

You lose 4 apples, leaving you with 6 apples.
You gain a gold coin, so now you have 4 coins (3 + 1).

Then, three birds drop 6 apples each, which is a total of 18 apples. You now have:
6 (initial apples) + 18 (new apples) = 24 apples

In the online game, you win 6 gold coins, but since you have to share with teammates, each of you gets 2 coins (6 ÷
3). So, you gain 2 new coins.

You now have a total of 6 coins (4 initial + 2 gained).

With your 6 coins, you can buy apples at a rate of 0.5 coins per apple.
To find out how many apples you can buy, divide the number of coins by the price per apple:
6 coins ÷ 0.5 coins/apple = 12 apples

So, after buying apples with all your coins, you have a total of 24 (initial) + 12 (new) = 36 apples.

As for where the river is, it runs near a big city that has something to do with spending gold coins. Since you 
bought apples with your coins, we can infer that this city might be a market or a place where fruits and goods are 
sold. However, the exact location of the river is not specified in the story.

>>> Two years ago, I was three times as old as my brother was. In three years, I will be twice as old as my brother. How 
... old is each of us now? 
... 
Let's use variables to represent the current ages:

* My current age = M
* My brother's current age = B

Two years ago, the equation would be:

M - 2 = 3(B - 2)

Since I was three times as old as my brother two years ago, we can simplify the equation:

M - 2 = 3B - 6
M = 3B - 4

Now, let's look at the future situation in three years:

M + 3 = 2(B + 3)

Expanding and simplifying:

M + 3 = 2B + 6
M = 2B + 3

We now have two equations with two variables:

1) M = 3B - 4
2) M = 2B + 3

Equating the right sides of these equations, we get:

3B - 4 = 2B + 3

Simplifying and solving for B (my brother's current age):

B = 7

Now that we know my brother's current age, we can find mine:

M = 3B - 4
= 3(7) - 4
= 21 - 4
= 17

So, I am currently 17 years old, and my brother is currently 7 years old.

🧩 Configuration

models:
    - model: cstr/llama3-8b-spaetzle-v33
    - model: cstr/llama3-8b-spaetzle-v38
    - model: cstr/llama3-8b-spaetzle-v36
    - model: cstr/llama3-8b-spaetzle-v28
merge_method: model_stock
base_model: cstr/llama3-8b-spaetzle-v33
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "cstr/llama3-8b-spaetzle-v39"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])

EU AI Act Art. 53 — provider obligations

Added 2026-08-02 during an account-wide provenance review.

This is a model merge, not a format conversion. Most cstr/* repositories are GGUF conversions, where the upstream research team remains the provider of the model and the conversion changes only the numeric representation of the weights. A merge produces a model that did not previously exist, so under Regulation (EU) 2024/1689 the maintainer of this repository is plausibly the provider of it, and the duties that survive the Art. 53(2) free-and-open-source exemption — Art. 53(1)(c) and 53(1)(d) — attach here rather than upstream.

Art. 53(1)(c) — copyright policy. No training corpus was assembled by this repository. Merging combines weights that other providers already published; it performs no text or data mining, so no rights reservation under Art. 4(3) of Directive (EU) 2019/790 was engaged by this step. Copyright questions arising from how the constituent models were themselves trained attach to their respective providers. Any credible claim that this repository redistributes material it has no right to redistribute will be acted on — contact via the Community tab.

Art. 53(1)(d) — training content. No data was used to train this model: it is a weight-space combination of models trained by others, and its training content is theirs. This card does not name its constituents in machine-readable frontmatter. Where a mergekit configuration is reproduced above, that configuration is the record of what went in.