Yi-34B

34.4B parameters, text in, text out, 4K tokens of context.

Verified 2026-10-25 Rank #451 in LLMs · score 78 method v1 Sources 3 Downloads 30d 10.3k · #503 of 619 tracked

Can I run it?21GB at INT4Estimatefits a 24 GB cardmethod

Can I use it?Apache 2.0Permissive; patent grantlicence file2026-10-24

Do people use it?10.3kdownloads, 30 days#503 of 619 we track · below the median+4 likes, downloads +1% this week1,314 likesHugging Face2026-10-24

Worth reading

Picked by rule from 7 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Most discussed · 41 HN pts

Yi-34B, Llama 2, and common practices in LLM training

Yi-34B adopts Llama 2's architecture but utilizes standard techniques like the SwiGLU Activation Function and Rotary Positional Embeddings (RoPE) that were publicly known before Meta's release.

Grouped-Query Attention was proposed relatively recently in 2023 by Google Research as an extension of Multi-Query Attention.
Training a language model requires assembling a dataset and building training infrastructure on one's own computing clusters with high performance computing expertise.

blog.eleuther.ai 2024-04-04 41 HN pts · 3 comments 1.7k words

Latest · 3 years ago

SHOW HN:A New 34B Open Source LLM, Astonishing 78 Score in MMLU (GPT-4 MMLU:83)

OrionStar-Yi-34B-Chat is a fine-tuned version of the open-source Yi-34B model trained on a high-quality corpus of over 150k sentences.

The model is loaded with device_map='auto' to utilize all available GPUs during inference.
The command-line chat tool does not support calling the base model.

github.com 2023-11-20 21 HN pts · 5 comments 923 words

7 best of 7 tracked

The 3 picks above, plus these 4. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

ArticleYi-Coder: A Small but Mighty LLM for Code01-ai.github.io · 2024-09-05
DiscussionLocal LLMold.reddit.com · 2023-11-15 · 10 HN pts
DiscussionNous-Hermes-2-Yi-34Bold.reddit.com · 2023-12-26 · 1 HN pts
Article4 Command lines to run open source LLMs across devices with 2MB inference appsecondstate.io · 2023-11-21 · 1 HN pts

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

No API provider serves it, so its downloads are people running it themselves: 9 quantized builds and 81 Spaces.

As a service0no API provider listed
On own hardware9quantized builds
Built on1711 fine-tunes · 3 adapters · 3 merges
In apps81Hugging Face Spaces

Top Spaces running it: KBaba7/Quant ♥ 51 · allenai/URIAL-Bench ♥ 22 · FallnAI/Quantize-HF-Models ♥ 19 · bhaskartripathi/LLM_Quantization ♥ 16

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits a 64 GB Mac (unified memory) at INT4

About 21 GB at INT4, 41 GB at FP8, 83 GB at BF16: weights only, so leave room for the KV cache.

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

6 dated articles and 30 repo commits so far

Teal is the weights (2023-11-01), amber triangles are 6 dated articles (each links to it), grey ticks are repo commits.

What people built on it

26 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

26 derivatives on the Hub, led by 11 fine-tunes

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF mradermacher/Yi-34B-i1-GGUF for llama.cpp, Ollama, LM Studio 79.8k downloads MLX k8smee/Yi-34B-mlx-6Bit for Apple silicon 293 downloads AWQ TheBloke/Yi-34B-AWQ for vLLM, SGLang 13.4k downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Facts

21 fields

Listed in LLMs

Facts

Yi-34B has 34.4B parameters. (safetensors metadata, observed 2026-10-24)
Context length is 4,096 tokens (4K). (config.json max_position_embeddings, observed 2026-10-24)
License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
Created 2023-11-01, last modified 2024-11-11. (HF API, observed 2026-10-24)

Pipeline tag	text-generation
Library	transformers
Architectures	LlamaForCausalLM
Model type	llama
Dtype	bfloat16
Params by dtype	BF16: 34,388,917,248
Size band	L
MoE	no
Context length	4,096 (config.json max_position_embeddings)
Papers	arXiv:2403.04652, arXiv:2311.16502, arXiv:2401.11944
License	Apache 2.0 (apache-2.0)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · LlamaForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · LlamaForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for Yi-34B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	83 GB · Multi-GPU node (<640 GB)
FP8 (1 byte/param)	41 GB · Single high-end GPU (<80 GB)
INT4 (0.5 bytes/param)	21 GB · Single GPU (<24 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

All variants

26 variants

Lineage

Original model Family: 01-ai/Yi 26 derivatives on the Hub

finetune: 11quantized: 9merge: 3adapter: 3

Top variants (26 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
NousResearch/Nous-Hermes-2-Yi-34B	finetune	—	34.4B	L	apache-2.0	NousResearch	266	551.5k
mradermacher/Yi-34B-i1-GGUF	quantized	GGUF	—	—	apache-2.0	mradermacher	1	79.8k
mradermacher/Yi-34B-GGUF	quantized	GGUF	—	—	apache-2.0	mradermacher	0	60.2k
NousResearch/Nous-Hermes-2-Yi-34B-GGUF	quantized	GGUF	—	—	apache-2.0	NousResearch	54	55.3k
TheBloke/Yi-34B-GGUF	quantized	GGUF	—	—	other	TheBloke	77	34.2k
TheBloke/Yi-34B-AWQ	quantized	AWQ	34.4B	L	other	TheBloke	7	13.4k
TheBloke/Yi-34B-GPTQ	quantized	GPTQ	34.4B	L	other	TheBloke	32	3.5k
tensorblock/Yi-34B-GGUF	quantized	GGUF	—	—	apache-2.0	tensorblock	0	1.8k
patf82/Nous-Hermes-2-Yi-34B-IQ3-imatrix-GGUF	quantized	imatrix	—	—	—	patf82	1	986
LoneStriker/Nous-Hermes-2-Yi-34B-4.0bpw-h6-exl2	finetune	0bpw	—	—	apache-2.0	LoneStriker	5	376
LoneStriker/Nous-Hermes-2-Yi-34B-4.65bpw-h6-exl2	finetune	5bpw	—	—	apache-2.0	LoneStriker	1	276
LoneStriker/Nous-Hermes-2-Yi-34B-5.0bpw-h6-exl2	finetune	0bpw	—	—	apache-2.0	LoneStriker	1	253
k8smee/Yi-34B-mlx-6Bit	quantized	mlx	34.4B	L	apache-2.0	k8smee	0	293
LoneStriker/Nous-Hermes-2-Yi-34B-6.0bpw-h6-exl2	finetune	0bpw	—	—	apache-2.0	LoneStriker	0	240
LoneStriker/Nous-Hermes-2-Yi-34B-3.0bpw-h6-exl2	finetune	0bpw	—	—	apache-2.0	LoneStriker	2	71
bartowski/Nous-Hermes-2-Yi-34B-exl2	finetune	exl2	—	—	apache-2.0	bartowski	1	95
LoneStriker/Nous-Hermes-2-Yi-34B-8.0bpw-h8-exl2	finetune	0bpw	—	—	apache-2.0	LoneStriker	0	76
mergekit-community/mergekit-slerp-epibiuy	merge	—	34.4B	L	—	mergekit-community	0	70
mergekit-community/mergekit-ties-hwemsto	merge	—	34.4B	L	—	mergekit-community	0	69
hiiamsid/yi_4k_paged_optimizer	finetune	—	34.4B	L	other	hiiamsid	0	68
Kardanil/leagaleasy-01-ai-Yi-34B	adapter	—	—	—	other	Kardanil	0	64
hiiamsid/yi_34B_8k_classification	finetune	—	34.4B	L	other	hiiamsid	0	54
Shaleen123/yi-34b-medical-full	adapter	—	—	—	—	Shaleen123	0	37
soyk/cs_address_parser	adapter	—	—	—	—	soyk	0	32
soyk/yi-32b-lb	finetune	—	—	—	other	soyk	0	0
Abigail45/Chyio	merge	—	—	—	apache-2.0	Abigail45	0	0

Repository changelog

30 commits

Repository changelog 30 commits

Date	Commit	SHA
2024-11-11	Auto Sync from git://github.com/01-ai/Yi.git/commit/c86f355fefb0b3965e7b29dc4265c1fe7eb28662	e1fa7c8
2024-11-05	Auto Sync from git://github.com/01-ai/Yi.git/commit/d4f0881d328643d1d3e85b52f9de4efce40a3895	252866c
2024-07-29	Auto Sync from git://github.com/01-ai/Yi.git/commit/09d246c97195ac87a40d0be35be48cb632e694ea	e7c4235
2024-07-26	Auto Sync from git://github.com/01-ai/Yi.git/commit/50bf141201486a3d3833dce66c07ab1933c196f9	e489021
2024-06-26	Auto Sync from git://github.com/01-ai/Yi.git/commit/11d140d641d18e1190caad2d5170da3b76c9e4f6	bc84a5d
2024-06-11	Auto Sync from git://github.com/01-ai/Yi.git/commit/3fceb848f1b996978b5ee718492031ad7576fea5	e1e7da8
2024-05-30	Auto Sync from git://github.com/01-ai/Yi.git/commit/7590b4438733986ef3d410e5b0a3241e430c19ac	7d30c79
2024-05-30	Auto Sync from git://github.com/01-ai/Yi.git/commit/9038a0a485b3e82ff841cc1bbb9d84f7b75fa421	d86275b
2024-05-29	Auto Sync from git://github.com/01-ai/Yi.git/commit/424329fdcf69b091cde1891dc2424b206f3573b3	da71e56
2024-04-29	Auto Sync from git://github.com/01-ai/Yi.git/commit/845a0ffb37528402199c68687ecbbfb675b039df	51a24ad
2024-04-26	Auto Sync from git://github.com/01-ai/Yi.git/commit/517b411e18f4070c56c6ef43a9dc977a709d3b26	4d0d5af
2024-04-24	Auto Sync from git://github.com/01-ai/Yi.git/commit/d89f53c4ca23bbaa062ebe1ab7f2aeff54db8aee	630735e
2024-04-24	Auto Sync from git://github.com/01-ai/Yi.git/commit/87ab46f0f3e054c7d161ece1492c77a1789d8b8a	8200882
2024-04-18	Auto Sync from git://github.com/01-ai/Yi.git/commit/25f046b33748f9ad4a49c227307dcaffee39af34	f9cec17
2024-04-17	Auto Sync from git://github.com/01-ai/Yi.git/commit/731b2af8583cba38d6544ebf909d7c85545f75a8	faac868
2024-04-08	Auto Sync from git://github.com/01-ai/Yi.git/commit/b10bf347b810b6ab1a7e73a7e1a1223c757e7326	629dd3c
2024-03-26	Auto Sync from git://github.com/01-ai/Yi.git/commit/5ed73fe931e7b95ad6dbd4d8091cd21b199d9bbb	533e00c
2024-03-22	Auto Sync from git://github.com/01-ai/Yi.git/commit/e3e47d27d65641181e2aeb600304181dd2a7230a	c49309d
2024-03-21	Auto Sync from git://github.com/01-ai/Yi.git/commit/704d5c148e087e9d1c83fb51e02790b197ce1aba	951e3f5
2024-03-20	Auto Sync from git://github.com/01-ai/Yi.git/commit/c3a9089c9b1cb42de485c1f7e6b36b50fa5763fe	7e1fefc
2024-03-20	Auto Sync from git://github.com/01-ai/Yi.git/commit/5e1c29a4f2224ca3ba1043347324d9b94f9737f6	ecaa9c2
2024-03-16	Auto Sync from git://github.com/01-ai/Yi.git/commit/b5d6c602a213bb9fd4832b188d1457a7da77a09c	678f15b
2024-03-08	Update README.md (#47)	2c95320
2024-03-08	Auto Sync from git://github.com/01-ai/Yi.git/commit/57b9e9f4e777740875aae331382d394997a97513	cde5a0e
2024-03-08	Auto Sync from git://github.com/01-ai/Yi.git/commit/b4528f877b729e2d976e711e35a21a8d112019d4	4e005f8
2024-03-07	Auto Sync from git://github.com/01-ai/Yi.git/commit/6161f43679e3887a500a4f00aec3cb3090e3a23a	62711dc
2024-03-07	Auto Sync from git://github.com/01-ai/Yi.git/commit/7acc118e8ab00dbf223af5549efdf0e3ef972701	c00ce7b
2024-03-06	Auto Sync from git://github.com/01-ai/Yi.git/commit/81dbb7886c95ed0754a6b62805e8fa7bc7db60d8	36be0a7
2024-03-06	Auto Sync from git://github.com/01-ai/Yi.git/commit/6da0100d379e4446978ab09f10fd58c6cbdc452d	1204768
2024-03-06	Auto Sync from git://github.com/01-ai/Yi.git/commit/971fab7959bbae9113729c5d77f10c3d8dc1e00c	535af03

Articles and coverage

7 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official1

Yi-34B, Llama 2, and common practices in LLM training

blog.eleuther.ai2024-04-04HN 41 pts · 3 comments1,675 words

Yi-34B adopts Llama 2's architecture but utilizes standard techniques like the SwiGLU Activation Function and Rotary Positional Embeddings (RoPE) that were publicly known before Meta's release. The primary differentiator remains the training dataset, as 01.AI developed and described their own English-Chinese dataset to ensure performance in Chinese. This addresses a November 2023 Hugging Face issue regarding component naming conventions rather than model dependency.written, reviewed, cited

Grouped-Query Attention was proposed relatively recently in 2023 by Google Research as an extension of Multi-Query Attention.
Training a language model requires assembling a dataset and building training infrastructure on one's own computing clusters with high performance computing expertise.

Analysis2

SHOW HN:A New 34B Open Source LLM, Astonishing 78 Score in MMLU (GPT-4 MMLU:83)

github.com2023-11-20authorityHN 21 pts · 5 comments923 words

OrionStar-Yi-34B-Chat is a fine-tuned version of the open-source Yi-34B model trained on a high-quality corpus of over 150k sentences. The model achieves 77.71 on C-Eval and 78.32 on MMLU while adhering to the Apache 2.0 and Yi-34B Model Community License Agreement.written, reviewed, cited

The model is loaded with device_map='auto' to utilize all available GPUs during inference.
The command-line chat tool does not support calling the base model.
Users are urged not to use the model for internet services without proper security review and filing.

4 Command lines to run open source LLMs across devices with 2MB inference app

secondstate.io2023-11-21HN 1 pts · 0 comments529 words

The dolphin-2.2-yi-34b model is a fine-tuned version of the Yi-34B LLM converted to llama2 format and further optimized for local deployment. It can be run on personal devices using a Rust and Wasm stack without requiring complex Python packages or C++ toolchains. On a Jetson Orin device, the model clocks in at about 7 tokens per second.written, reviewed, cited

Community3

Local LLM

old.reddit.com2023-11-15about the Yi familyHN 10 pts · 0 comments

AI cypher match: ChatGPT and Yi-34B play a game of hidden codes

twitter.com2023-12-12HN 1 pts · 0 comments

Nous-Hermes-2-Yi-34B

old.reddit.com2023-12-26HN 1 pts · 0 comments

All coverage tracked for this model (7) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
Yi-Coder: A Small but Mighty LLM for Code	01-ai.github.io	2024-09-05	2026-10-07	hn
Yi-34B, Llama 2, and common practices in LLM training	blog.eleuther.ai	2024-04-04	2026-10-07	hn
Nous-Hermes-2-Yi-34B	old.reddit.com	2023-12-26	2026-10-07	hn
AI cypher match: ChatGPT and Yi-34B play a game of hidden codes	twitter.com	2023-12-12	2026-10-07	hn
4 Command lines to run open source LLMs across devices with 2MB inference app	secondstate.io	2023-11-21	2026-10-07	hn
SHOW HN:A New 34B Open Source LLM, Astonishing 78 Score in MMLU (GPT-4 MMLU:83)	github.com	2023-11-20	2026-10-07	hn
Local LLM	old.reddit.com	2023-11-15	2026-10-07	hn

Papers

3 papers

Papers 3

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

Yi: Open Foundation Models by 01.AI cited in card

01. AI, Alex Young, Bei Chen +29 · 2024-03-07 · PDF · HF paper page (65 upvotes)
MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI cited in card

Xiang Yue, Yuansheng Ni, Kai Zhang +19 · 2023-11-27 · PDF · HF paper page (40 upvotes)
CMMMU: A Chinese Massive Multi-discipline Multimodal Understanding Benchmark cited in card

Ge Zhang, Xinrun Du, Bei Chen +19 · 2024-01-22 · PDF · HF paper page (27 upvotes)

Similar models

3 models

Similar models

Qwen3-32B · 32.8B
same size band (L), same task (text-generation), more providers, permissive license
Qwen2.5-Coder-32B-Instruct · 32.8B
same size band (L), same task (text-generation), more providers, permissive license
DeepSeek-R1-Distill-Qwen-32B · 32.8B
same size band (L), same task (text-generation), more providers, permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:40 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:12 UTC · Last verified 2026-10-25

Follow Yi-34B

Watch this model

/feeds/models/01-ai--Yi-34B.xml(RSS 2.0) · every change to Yi-34B
/feeds/families/01-ai--Yi.xml(RSS 2.0) · every change across the Yi family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.