Name: Bagel 34B V0.2
Author: jondurbin

Bagel 34B V0.2 is an open-source language model by jondurbin. Features: 34b LLM, VRAM: 68.7GB, Context: 195K, License: apache-2.0, HF Score: 69.7, LLM Explorer Score: 0.22, Arc: 68.8, HellaSwag: 83.7, MMLU: 76.5, TruthfulQA: 59.3, WinoGrande: 83.8, GSM8K: 46.2.

Autotrain compatible Conversational Dataset:ai2 arc Dataset:allenai/ultrafeedback ... Dataset:boolq Dataset:cais/mmlu Dataset:cakiki/rosetta-code Dataset:codeparrot/apps Dataset:datasets/winogrande Dataset:drop Dataset:facebook/belebele Dataset:intel/orca dpo pairs Dataset:jondurbin/cinematika-v... Dataset:jondurbin/truthy-dpo-v... Dataset:julielab/emobank Dataset:kingbri/pippa-sharegpt Dataset:ldjnr/capybara Dataset:lmsys/lmsys-chat-1m Dataset:migtissera/synthia-v1.... Dataset:muennighoff/natural-in... Dataset:nvidia/helpsteer Dataset:open-orca/slimorca Dataset:openbookqa Dataset:piqa Dataset:spider Dataset:squad v2 Dataset:squish42/bluemoon-fand... Dataset:tiger-lab/mathinstruct Dataset:unalignment/spicy-3.1 Dataset:unalignment/toxic-dpo-... Dataset:vezora/tested-22k-pyth... Endpoints compatible Llama Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/jondurbin/bagel-34b-v0.2

Bagel 34B V0.2 Benchmarks

IFEval: 45.56 vs 88 (so35)^-48.2%

ARC: 68.77 vs 96.7 (so35)^-28.9%

HellaSwag: 83.72 vs 95.3 (gpt4)^-12.2%

MMLU: 76.45 vs 88.3 (so35)^-13.4%

TruthfulQA: 59.26 vs 59 (gpt4)^0.4%

WinoGrande: 83.82 vs 87.5 (gpt4)^-4.2%

GSM8K: 46.17 vs 96.4 (so35)^-52.1%

MATH Lvl 5: 0.15

LLME Score: 0.22147

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

Bagel 34B V0.2 (jondurbin/bagel-34b-v0.2)

🌟 Advertise your project 🚀

Bagel 34B V0.2 Parameters and Internals

Model Type

Text Generation

Training Details

Data Sources:

ai2_arc, unalignment/spicy-3.1, codeparrot/apps, facebook/belebele, boolq, jondurbin/cinematika-v0.1, drop, lmsys/lmsys-chat-1m, TIGER-Lab/MathInstruct, cais/mmlu, Muennighoff/natural-instructions, openbookqa, piqa, Vezora/Tested-22k-Python-Alpaca, cakiki/rosetta-code, Open-Orca/SlimOrca, spider, squad_v2, migtissera/Synthia-v1.3, datasets/winogrande, nvidia/HelpSteer, Intel/orca_dpo_pairs, unalignment/toxic-dpo-v0.1, jondurbin/truthy-dpo-v0.1, allenai/ultrafeedback_binarized_cleaned, Squish42/bluemoon-fandom-1-1-rp-cleaned, LDJnr/Capybara, JULIELab/EmoBank, kingbri/PIPPA-shareGPT

Methodology:

SFT phase using a variety of prompt formats: vicuna, llama-2, alpaca, and chat-ml (sorta).

Model Architecture:

unknown

Input Output

Input Format:

vicuna, llama-2, alpaca, chat-ml

Accepted Modalities:

text

Output Format:

text

Performance Tips:

Recommended to run for 1 epoch or 0.75 epoch, using a relatively low learning rate.

LLM Name	Bagel 34B V0.2
Repository 🤗	https://huggingface.co/jondurbin/bagel-34b-v0.2
Model Size	34b
Required VRAM	68.7 GB
Updated	2025-11-08
Maintainer	jondurbin
Model Type	llama
Model Files	4.0 GB: 1-of-18 3.9 GB: 2-of-18 3.9 GB: 3-of-18 3.9 GB: 4-of-18 3.9 GB: 5-of-18 3.9 GB: 6-of-18 3.9 GB: 7-of-18 3.9 GB: 8-of-18 3.9 GB: 9-of-18 3.9 GB: 10-of-18 3.9 GB: 11-of-18 3.9 GB: 12-of-18 3.9 GB: 13-of-18 3.9 GB: 14-of-18 3.9 GB: 15-of-18 3.9 GB: 16-of-18 3.9 GB: 17-of-18 2.3 GB: 18-of-18
Model Architecture	LlamaForCausalLM
License	apache-2.0
Context Length	200000
Model Max Length	200000
Transformers Version	4.36.2
Tokenizer Class	LlamaTokenizer
Padding Token	<unk>
Vocabulary Size	64000
Torch Data Type	bfloat16

Quantized Models of the Bagel 34B V0.2

Model	Likes	Downloads	VRAM
Luminex 34B V0.1 EXL2 8.0bpw	1	1	34 GB
Bagel 34B V0.2 GGUF	3	676	14 GB
Smaug 34B V0.1 2.65bpw H6 EXL2	1	7	12 GB
Smaug 34B V0.1 4.0bpw H6 EXL2	3	4	18 GB
Smaug 34B V0.1 3.0bpw H6 EXL2	2	5	13 GB
Smaug 34B V0.1 4.65bpw H6 EXL2	1	9	20 GB
Smaug 34B V0.1 2.4bpw H6 EXL2	1	0	11 GB
Smaug 34B V0.1 GPTQ	1	8	21 GB
Smaug 34B V0.1 8.0bpw H8 EXL2	1	5	34 GB
Smaug 34B V0.1 AWQ	2	0	19 GB

Best Alternatives to Bagel 34B V0.2

Best Alternatives	Context / RAM	Downloads	Likes
Yi 34B 200K	195K / 68.9 GB	11042	321
Bagel Hermes 34B Slerp	195K / 68.9 GB	8712	1
34B Beta	195K / 69.2 GB	7868	64
Smaug 34B V0.1	195K / 69.2 GB	8634	61
Casual Magnum 34B	195K / 68.8 GB	5	1
Yi 34B 200K AEZAKMI V2	195K / 69.2 GB	1872	12
Faro Yi 34B	195K / 69.2 GB	8679	6
Bagel DPO 34B V0.5	195K / 68.7 GB	8055	17
Smaug 34B V0.1 ExPO	195K / 69.2 GB	7108	0
Luminex 34B V0.1	195K / 68.9 GB	9240	8

Note: green Score (e.g. "73.2") means that the model is better than jondurbin/bagel-34b-v0.2.

Rank the Bagel 34B V0.2 Capabilities

🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

What open-source LLMs or SLMs are you in search of? 51638 in total.

Email us: info@extractum.io. Our Privacy Policy | Terms and Conditions | Suggest an improvement.

Our Social Media →

Original data from HuggingFace, OpenCompass and various public git repos.

Check out Ag3ntum — our secure, self-hosted AI agent for server management.

Release v20241124

Support LLM Explorer

Bagel 34B V0.2 by jondurbin

» All LLMs » jondurbin » Bagel 34B V0.2 URL Share it on