Spaetzle-v85-7b

Name: Spaetzle-v85-7b
Rating: 2.5 (17 reviews)

Spaetzle-v85-7b is a merge of the following models using LazyMergekit:

Evaluation

EQ-Bench (v2_de): 65.32, Parseable: 171.0

Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
Spaetzle-v85-7b	44.35	75.99	67.23	46.55	58.53

From Intel/low_bit_open_llm_leaderboard:

Metric	Value
ARC-c	62.63
ARC-e	85.56
Boolq	87.77
HellaSwag	66.66
Lambada	70.35
MMLU	61.61
Openbookqa	37.2
Piqa	82.48
Truthfulqa	50.43
Winogrande	78.3
Average	68.3

From Occiglot Euro LLM Leaderboard

Model	🇪🇺 Average ⬆️	🇩🇪 DE	🇬🇧 EN	🇬🇧ARC EN	🇬🇧TruthfulQA EN	🇬🇧Belebele EN	🇬🇧HellaSwag EN	🇬🇧MMLU EN	🇩🇪ARC DE	🇩🇪TruthfulQA DE	🇩🇪Belebele DE	🇩🇪HellaSwag DE	🇩🇪MMLU DE
mistral-community/Mixtral-8x22B-v0.1	68.3	66.81	72.87	70.56	52.29	93.89	70.41	77.17	63.9	29.31	92.44	77.9	70.49
cstr/Spaetzle-v85-7b	63.26	61.11	71.94	70.48	67.16	90.33	68.54	63.17	58.43	36.93	84.22	70.62	55.36
cstr/Spaetzle-v60-7b	63.32	60.95	71.65	69.88	66.24	90.11	68.43	63.59	58	37.31	84.22	70.09	55.11
VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct	64.49	60.07	74.71	74.49	66.19	91.67	74.55	66.65	59.37	29.57	88.56	66.43	56.44
seedboxai/Llama-3-KafkaLM-8B-v0.1	62.27	59.67	69.75	69.03	58.14	90.78	64.35	66.43	57.66	30.33	85.89	66.88	57.58
cstr/llama3-8b-spaetzle-v33	62.75	59.56	70.68	69.54	59.31	91.44	66.04	67.06	57.06	28.55	87.56	66.7	57.92

AGIEval

Task	Version	Metric	Value		Stderr
agieval_aqua_rat	0	acc	23.23	±	2.65
		acc_norm	22.44	±	2.62
agieval_logiqa_en	0	acc	37.33	±	1.90
		acc_norm	37.94	±	1.90
agieval_lsat_ar	0	acc	25.22	±	2.87
		acc_norm	23.04	±	2.78
agieval_lsat_lr	0	acc	49.41	±	2.22
		acc_norm	50.78	±	2.22
agieval_lsat_rc	0	acc	64.68	±	2.92
		acc_norm	63.20	±	2.95
agieval_sat_en	0	acc	77.67	±	2.91
		acc_norm	78.16	±	2.89
agieval_sat_en_without_passage	0	acc	46.12	±	3.48
		acc_norm	45.15	±	3.48
agieval_sat_math	0	acc	35.45	±	3.23
		acc_norm	34.09	±	3.20

Average: 44.35%

GPT4All

Task	Version	Metric	Value		Stderr
arc_challenge	0	acc	63.82	±	1.40
		acc_norm	64.76	±	1.40
arc_easy	0	acc	85.90	±	0.71
		acc_norm	82.32	±	0.78
boolq	1	acc	87.61	±	0.58
hellaswag	0	acc	67.39	±	0.47
		acc_norm	85.36	±	0.35
openbookqa	0	acc	38.80	±	2.18
		acc_norm	48.80	±	2.24
piqa	0	acc	83.03	±	0.88
		acc_norm	84.17	±	0.85
winogrande	0	acc	78.93	±	1.15

Average: 75.99%

TruthfulQA

Task	Version	Metric	Value		Stderr
truthfulqa_mc	1	mc1	50.80	±	1.75
		mc2	67.23	±	1.49

Average: 67.23%

Bigbench

Task	Version	Metric	Value		Stderr
bigbench_causal_judgement	0	multiple_choice_grade	54.74	±	3.62
bigbench_date_understanding	0	multiple_choice_grade	68.29	±	2.43
bigbench_disambiguation_qa	0	multiple_choice_grade	39.53	±	3.05
bigbench_geometric_shapes	0	multiple_choice_grade	22.28	±	2.20
		exact_str_match	12.26	±	1.73
bigbench_logical_deduction_five_objects	0	multiple_choice_grade	32.80	±	2.10
bigbench_logical_deduction_seven_objects	0	multiple_choice_grade	23.00	±	1.59
bigbench_logical_deduction_three_objects	0	multiple_choice_grade	59.00	±	2.84
bigbench_movie_recommendation	0	multiple_choice_grade	45.60	±	2.23
bigbench_navigate	0	multiple_choice_grade	51.10	±	1.58
bigbench_reasoning_about_colored_objects	0	multiple_choice_grade	70.10	±	1.02
bigbench_ruin_names	0	multiple_choice_grade	52.68	±	2.36
bigbench_salient_translation_error_detection	0	multiple_choice_grade	33.57	±	1.50
bigbench_snarks	0	multiple_choice_grade	71.27	±	3.37
bigbench_sports_understanding	0	multiple_choice_grade	74.54	±	1.39
bigbench_temporal_sequences	0	multiple_choice_grade	40.00	±	1.55
bigbench_tracking_shuffled_objects_five_objects	0	multiple_choice_grade	21.52	±	1.16
bigbench_tracking_shuffled_objects_seven_objects	0	multiple_choice_grade	18.86	±	0.94
bigbench_tracking_shuffled_objects_three_objects	0	multiple_choice_grade	59.00	±	2.84

Average: 46.55%

Average score: 58.53%

🧩 Configuration

yaml

models:
  - model: cstr/Spaetzle-v84-7b
    # no parameters necessary for base model
  - model: cstr/Spaetzle-v80-7b
    parameters:
      density: 0.65
      weight: 0.2
  - model: cstr/Spaetzle-v79-7b
    parameters:
      density: 0.65
      weight: 0.2
  - model: cstr/Spaetzle-v81-7b
    parameters:
      density: 0.65
      weight: 0.2
  - model: cstr/Spaetzle-v71-7b
    parameters:
      density: 0.65
      weight: 0.2
merge_method: dare_ties
base_model: cstr/Spaetzle-v84-7b
parameters:
  int8_mask: true
dtype: bfloat16
random_seed: 0
tokenizer_source: base

💻 Usage

python

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "cstr/Spaetzle-v85-7b"
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. Of the 4 constituent models this card names, 0 are still published and 4 are not: cstr/Spaetzle-v80-7b, cstr/Spaetzle-v79-7b, cstr/Spaetzle-v81-7b, cstr/Spaetzle-v71-7b. For those, the training-content chain cannot be followed from this card, and no summary is reconstructed here in their place — an untraceable summary presented as a traceable one would be worse than the gap.