List of large language models

For the training cost column, 1 petaFLOP-day equals 1 petaFLOP/sec × 1 day, or 8.64×10¹⁹ FLOP (floating point operations). Only the cost of the largest model is shown.

More information Name, Release date ...

Name	Release date[a]	Developer	Number of parameters (billion) [b]	Corpus size	Training cost (petaFLOP-day)	License[c]	Notes
GPT-1	June 11, 2018	OpenAI	0.117	Unknown	1[1]	MIT[2]	First GPT model, decoder-only transformer. Trained for 30 days on 8 P600 GPUs.[3]
BERT	October 2018	Google	0.340[4]	3.3 billion words[4]	9[5]	Apache 2.0[6]	An early and influential language model.[7] Encoder-only and thus not built to be prompted or generative.[8] Training took 4 days on 64 TPUv2 chips.[9]
T5	October 2019	Google	11[10]	34 billion tokens[10]		Apache 2.0[11]	Base model for many Google projects, such as Imagen.[12]
XLNet	June 2019	Google	0.340[13]	33 billion words	330	Apache 2.0[14]	An alternative to BERT; designed as encoder-only. Trained on 512 TPU v3 chips for 5.5 days.[15]
GPT-2	February 2019	OpenAI	1.5[16]	40GB[17] (~10 billion tokens)[18]	28[19]	MIT[20]	Trained on 32 TPUv3 chips for 1 week.[19]
GPT-3	May 2020	OpenAI	175[21]	300 billion tokens[18]	3640[22]	Proprietary	A fine-tuned variant of GPT-3, termed GPT-3.5, was made available to the public through a web interface called ChatGPT in 2022.[23]
GPT-Neo	March 2021	EleutherAI	2.7[24]	825 GiB[25]	Unknown	MIT[26]	The first of a series of free GPT-3 alternatives released by EleutherAI. GPT-Neo outperformed an equivalent-size GPT-3 model on some benchmarks, but was significantly worse than the largest GPT-3.[26]
GPT-J	June 2021	EleutherAI	6[27]	825 GiB[25]	200[28]	Apache 2.0	GPT-3-style language model trained on The Pile.
Megatron-Turing NLG	October 2021[29]	Microsoft and Nvidia	530[30]	338.6 billion tokens[30]	38000[31]	Unreleased	Trained for 3 months on over 2000 A100 GPUs on the NVIDIA Selene Supercomputer, for over 3 million GPU-hours.[31]
Ernie 3.0 Titan	December 2021	Baidu	260[32]	4TB	Unknown	Proprietary	Chinese-language LLM initially used by Ernie Bot.
Claude[33]	December 2021	Anthropic	52[34]	400 billion tokens[34]	Unknown	Proprietary	Fine-tuned for desirable behavior in conversations.[35]
GLaM (Generalist Language Model)	December 2021	Google	1200[36]	1.6 trillion tokens[36]	5600[36]	Proprietary	Sparse mixture of experts (MoE) model, making it more expensive to train but cheaper to run inference compared to GPT-3.
Gopher	December 2021	Google DeepMind	280[37]	300 billion tokens[38]	5833[39]	Proprietary	Later developed into the Chinchilla model.
LaMDA (Language Models for Dialog Applications)	January 2022	Google	137[40]	1.56T words,[40] 168 billion tokens[38]	4110[41]	Proprietary	Specialized for response generation in conversations.
GPT-NeoX	February 2022	EleutherAI	20[42]	825 GiB[25]	740[28]	Apache 2.0	Based on the Megatron architecture.
Chinchilla	March 2022	Google DeepMind	70[43]	1.4 trillion tokens[43][38]	6805[39]	Proprietary	Reduced-parameter model trained on more data. Used in the Sparrow bot. Often cited for its neural scaling law.
PaLM (Pathways Language Model)	April 2022	Google	540[44]	768 billion tokens[43]	29,250[39]	Proprietary	Trained for ~60 days on ~6000 TPU v4 chips.[39]
OPT (Open Pretrained Transformer)	May 2022	Meta	175[45]	180 billion tokens[46]	310[28]	Non-commercial research[d]	GPT-3 architecture with some adaptations from Megatron. The training logbook written by the team was published.[47]
YaLM 100B	June 2022	Yandex	100[48]	1.7TB[48]	Unknown	Apache 2.0	English-Russian model based on Microsoft's Megatron-LM.
Minerva	June 2022	Google	540[49]	38.5B tokens from webpages filtered for mathematical content and from papers submitted to the arXiv preprint server[49]	Unknown	Proprietary	For solving "mathematical and scientific questions using step-by-step reasoning".[50] Initialized from PaLM models, then finetuned on mathematical and scientific data.
BLOOM	July 2022	Large collaboration led by Hugging Face	175[51]	350 billion tokens (1.6TB)[52]	Unknown	Responsible AI	Essentially GPT-3 but trained on a multi-lingual corpus (30% English excluding programming languages).
Galactica	November 2022	Meta	120	106 billion tokens[53]	Unknown	CC-BY-NC-4.0	Trained on scientific text and modalities.
AlexaTM (Teacher Models)	November 2022	Amazon	20[54]	1.3 trillion[55]	Unknown	Proprietary[56]	Uses a bidirectional sequence-to-sequence architecture.
Llama	February 2023	Meta AI	65[57]	1.4 trillion[57]	6300[58]	Non-commercial research[e]	Corpus has 20 languages. "Overtrained" (compared to the Chinchilla scaling law) for better performance with fewer parameters.[57][failed verification]
GPT-4	March 2023	OpenAI	Unknown[f] (According to rumors: 1760)[60]	Unknown	Unknown, estimated 230,000	Proprietary	Now available for all ChatGPT users; used in several products.
Cerebras-GPT	March 2023	Cerebras	13[61]		270[28]	Apache 2.0	Trained with the Chinchilla neural scaling formula.
Falcon	March 2023	Technology Innovation Institute	40[62]	1 trillion tokens, from RefinedWeb (filtered web text corpus)[63] plus some "curated corpora".[64]	2800[58]	Apache 2.0[65]
BloombergGPT	March 2023	Bloomberg L.P.	50	363 billion tokens from Bloomberg's proprietary data sources, plus 345 billion tokens from general purpose datasets[66]	Unknown	Unreleased	Designed for financial tasks.[66]
PanGu-Σ	March 2023	Huawei	1085	329 billion tokens[67]	Unknown	Proprietary
OpenAssistant[68]	March 2023	LAION	17	1.5 trillion tokens	Unknown	Apache 2.0	Trained on crowdsourced, open conversational data.
Jurassic-2[69][70]	March 2023	AI21 Labs	Unknown	Unknown	Unknown	Proprietary
PaLM 2 (Pathways Language Model 2)	May 2023	Google	340[71]	3.6 trillion tokens[71]	85,000[58]	Proprietary	Used in the Bard chatbot.[72]
YandexGPT	May 17, 2023	Yandex	Unknown	Unknown	Unknown	Proprietary	Used in the Alice chatbot.
Llama 2	July 2023	Meta AI	70[73]	2 trillion tokens[73]	21,000	Llama 2 license	Trained over 3.3 million GPU (A100) hours.[74]
Claude 2	July 2023	Anthropic	Unknown	Unknown	Unknown	Proprietary	Used in the Claude chatbot.[75]
Granite 13b	July 2023	IBM	Unknown	Unknown	Unknown	Proprietary	Used in IBM Watsonx.[76]
Mistral 7B	September 2023	Mistral AI	7.3[77]	Unknown	Unknown	Apache 2.0
YandexGPT 2	September 7, 2023	Yandex	Unknown	Unknown	Unknown	Proprietary	Used in the Alice chatbot.
Claude 2.1	November 2023	Anthropic	Unknown	Unknown	Unknown	Proprietary	Used in the Claude chatbot. Has a context window of 200,000 tokens, or ~500 pages.[78]
Grok 1[79]	November 2023	xAI	314	Unknown	Unknown	Apache 2.0	Used in the Grok chatbot. Grok 1 has a context length of 8,192 tokens and has access to X (Twitter).[80]
Gemini 1.0	December 2023	Google DeepMind	Unknown	Unknown	Unknown	Proprietary	Multimodal model, comes in three sizes. Used in the chatbot of the same name.[81]
Mixtral 8x7B	December 2023	Mistral AI	46.7	Unknown	Unknown	Apache 2.0	Outperforms GPT-3.5 and Llama 2 70B on many benchmarks.[82] Mixture of experts model, with 12.9 billion parameters activated per token.[83]
DeepSeek-LLM	November 29, 2023	DeepSeek	67	2T tokens[84]: table 2	12,000	DeepSeek License	Trained on English and Chinese text. Used 1024 training FLOPs for 67B model, 10b FLOPs for 7B.[84]: figure 5
Phi-2	December 2023	Microsoft	2.7	1.4T tokens	419[85]	MIT	Trained on real and synthetic "textbook-quality" data over 14 days on 96 A100 GPUs.[85]
Gemini 1.5	February 2024	Google DeepMind	Unknown	Unknown	Unknown	Proprietary	Multimodal model based on a MoE architecture. Context window above 1 million tokens.[86]
Gemini Ultra	February 2024	Google DeepMind	Unknown	Unknown	Unknown	Proprietary
Gemma	February 2024	Google DeepMind	7	6T tokens	Unknown	Gemma Terms of Use[87]
OLMo	February 2024	Allen Institute for AI	7[88]	2T tokens[89]	Unknown	Apache 2.0
Claude 3	March 2024	Anthropic	Unknown	Unknown	Unknown	Proprietary	Includes three models: Haiku, Sonnet, and Opus.[90]
DBRX	March 2024	Databricks and Mosaic ML	136	12T tokens	Unknown	Databricks Open Model License[91][92]	Training cost 10 million USD.[citation needed]
YandexGPT 3 Pro	March 28, 2024	Yandex	Unknown	Unknown	Unknown	Proprietary	Used in Alice chatbot.
Fugaku-LLM	May 2024	Fujitsu, Tokyo Institute of Technology, etc.	13	380B tokens	Unknown	Fugaku-LLM Terms of Use[93]	The largest model ever trained on CPU-only, on the Fugaku supercomputer.[94]
Chameleon	May 2024	Meta AI	34[95]	4.4 trillion	Unknown	Non-commercial research[96]
Mixtral 8x22B[97]	April 17, 2024	Mistral AI	141	Unknown	Unknown	Apache 2.0
Phi-3	April 23, 2024	Microsoft	14[98]	4.8T tokens[citation needed]	Unknown	MIT	Marketed by Microsoft as a "small language model".[99]
Granite Code Models	May 2024	IBM	Unknown	Unknown	Unknown	Apache 2.0
YandexGPT 3 Lite	May 28, 2024	Yandex	Unknown	Unknown	Unknown	Proprietary	Used in the Alice chatbot.
Qwen2	June 2024	Alibaba Cloud	72[100]	3T tokens	Unknown	Qwen License	Multiple sizes, the smallest being 0.5B.
DeepSeek-V2	June 2024	DeepSeek	236	8.1T tokens	28,000	DeepSeek License	1.4M hours on H800.[101]
Nemotron-4	June 2024	Nvidia	340	9T tokens	200,000	NVIDIA Open Model License[102][103]	Trained for 1 epoch. Trained on 6144 H100 GPUs between December 2023 and May 2024.[104][105]
Claude 3.5	June 2024	Anthropic	Unknown	Unknown	Unknown	Proprietary	Initially, only one model, Sonnet, was released.[106] In October 2024, Sonnet 3.5 was upgraded, and Haiku 3.5 became available.[107]
Llama 3.1	July 2024	Meta AI	405	15.6T tokens	440,000	Llama 3 license	405B version took 31 million hours on H100-80GB, at 3.8E25 FLOPs.[108][109]
Grok-2	August 14, 2024	xAI	Unknown	Unknown	Unknown	xAI Community License Agreement[110][111]	Originally closed-source, then re-released as "Grok 2.5" under a source-available license in August 2025.[112][113]
OpenAI o1	September 12, 2024	OpenAI	Unknown	Unknown	Unknown	Proprietary	First LLM described as a "reasoning model".[114][115][better source needed]
YandexGPT 4 Lite and Pro	October 24, 2024	Yandex	Unknown	Unknown	Unknown	Proprietary	Used in the Alice chatbot.
Mistral Large	November 2024	Mistral AI	123	Unknown	Unknown	Mistral Research License	Upgraded over time. The latest version is 24.11.[116]
Pixtral	November 2024	Mistral AI	123	Unknown	Unknown	Mistral Research License	Multimodal. There is also a 12B version which is under Apache 2 license.[116]
OLMo 2	November 2024	Allen Institute for AI	32[117][118]	6.6T tokens[118]	15,000[118]	Apache 2.0	Initially had 7B and 13B parameter variants, with 32B released later.
Phi-4	December 12, 2024	Microsoft	14[119]	9.8T tokens	Unknown	MIT	Marketed by Microsoft as a "small language model".[120]
DeepSeek-V3	December 2024	DeepSeek	671	14.8T tokens	56,000	MIT	Used 2.788M training hours on H800 GPUs.[121] Originally released under the DeepSeek License, then re-released under the MIT License as "DeepSeek-V3-0324" in March 2025.[122]
Amazon Nova	December 2024	Amazon	Unknown	Unknown	Unknown	Proprietary	Includes three models: Nova Micro, Nova Lite, and Nova Pro.[123]
DeepSeek-R1	January 2025	DeepSeek	671	Not applicable	Unknown	MIT	No pretraining; reinforcement-learned upon V3-Base.[124][125]
Qwen2.5	January 2025	Alibaba	72	18T tokens	Unknown	Qwen License	7 dense models with parameter counts from 0.5B to 72B. Alibaba also released 2 MoE variants.[126]
MiniMax-Text-01	January 2025	Minimax	456	4.7T tokens[127]	Unknown	Minimax Model license	[128][127]
Gemini 2.0	February 2025	Google DeepMind	Unknown	Unknown	Unknown	Proprietary	Three models released: Flash, Flash-Lite and Pro.[129][130][131]
Claude 3.7	February 24, 2025	Anthropic	Unknown	Unknown	Unknown	Proprietary	One model, Sonnet 3.7.[132]
YandexGPT 5 Lite Pretrain and Pro	February 25, 2025	Yandex	Unknown	Unknown	Unknown	Proprietary	Used in the Alice Neural Network chatbot.
GPT-4.5	February 27, 2025	OpenAI	Unknown	Unknown	Unknown	Proprietary	OpenAI's largest non-reasoning model at the time.[133]
Grok 3	February 2025	xAI	Unknown	Unknown	Unknown	Proprietary	Training cost claimed to be "10x the compute of previous state-of-the-art models".[134]
Gemini 2.5	March 25, 2025	Google DeepMind	Unknown	Unknown	Unknown	Proprietary	Three models released: Flash, Flash-Lite and Pro.[135]
YandexGPT 5 Lite Instruct	March 31, 2025	Yandex	Unknown	Unknown	Unknown	Proprietary	Used in the Alice Neural Network chatbot.
Llama 4	April 5, 2025	Meta AI	400	40T tokens	Unknown	Llama 4 license	[136][137]
OpenAI o3 and o4-mini	April 16, 2025	OpenAI	Unknown	Unknown	Unknown	Proprietary	Reasoning models.[138]
Qwen3	April 2025	Alibaba Cloud	235	36T tokens	Unknown	Apache 2.0	Multiple sizes, the smallest being 0.6B.[139]
Claude 4	May 22, 2025	Anthropic	Unknown	Unknown	Unknown	Proprietary	Includes two models, Sonnet and Opus.[140]
Sarvam-M	May 23, 2025	Sarvam AI	24	Unknown	Unknown	Apache 2.0	Hybrid reasoning model fine-tuned on Mistral Small base; optimized for math, programming, and Indian languages.[141][142]
Grok 4	July 9, 2025	xAI	Unknown	Unknown	Unknown	Proprietary	[citation needed]
Param-1	July 21, 2025	BharatGen	2.9[143]	5T tokens "focus[ed] on India’s linguistic landscape"[143]	Unknown	Unknown
GLM-4.5	July 29, 2025	Zhipu AI	355	22T tokens[144][g]	Unknown	MIT	Released in 335B and 106B sizes.[145]
GPT-OSS	August 5, 2025	OpenAI	117	Unknown	Unknown	Apache 2.0	Released in 20B and 120B sizes.[146]
Claude 4.1	August 5, 2025	Anthropic	Unknown	Unknown	Unknown	Proprietary	Includes one model, Opus.[147]
GPT-5	August 7, 2025	OpenAI	Unknown	Unknown	Unknown	Proprietary	Includes three models: GPT-5, GPT-5 mini, and GPT-5 nano. GPT-5 is available in ChatGPT and API. It includes reasoning abilities. [148][149]
DeepSeek-V3.1	August 21, 2025	DeepSeek	671	15.639T		MIT	Based on DeepSeek V3 (trained on 14.8T tokens); further trained on 839B tokens from the extension phases (630B + 209B).[150] A hybrid model that can switch between thinking and non-thinking modes.[151]
YandexGPT 5.1 Pro	August 28, 2025	Yandex	Unknown	Unknown	Unknown	Proprietary	Used in the Alice Neural Network chatbot.
Apertus	September 2, 2025	ETH Zurich and EPF Lausanne	70	15 trillion[152]	Unknown	Apache 2.0	The first LLM to be compliant with theArtificial Intelligence Act of the European Union.[153]
Claude Sonnet 4.5	September 29, 2025	Anthropic	Unknown	Unknown	Unknown	Proprietary	[154]
DeepSeek-V3.2-Exp	September 29, 2025	DeepSeek	685			MIT	Experimental model built upon v3.1-Terminus; uses a custom DeepSeek Sparse Attention (DSA) model.[155][156][157]
GLM-4.6	September 30, 2025	Zhipu AI	357			Apache 2.0	[158][159][160]
Alice AI LLM 1.0	October 28, 2025	Yandex	Unknown	Unknown	Unknown	Proprietary	Available in the Alice AI chatbot.
Gemini 3	November 18, 2025	Google DeepMind	Unknown	Unknown	Unknown	Proprietary	Two models released: Deep Think and Pro.[161]
Olmo 3[162]	November 20, 2025	Allen Institute for AI	32	5.9T tokens[163]	Unknown	Apache 2.0	Includes 7B and 32B parameter versions, alongside reasoning and instruction-following models.[163]
Claude Opus 4.5	November 24, 2025	Anthropic	Unknown	Unknown	Unknown	Proprietary	The largest model in the Claude family.[164]
GPT 5.2	December 11, 2025	OpenAI	Unknown	Unknown	Unknown	Proprietary	It was able to solve an open problem in statistical learning theory that had previously remained unresolved by human researchers.[165]
GLM-4.7	December 22, 2025	Zhipu AI	355			Apache 2.0	MoE architecture. Open-source state-of-the-art on coding benchmarks.[citation needed] A smaller Flash variant (30B-A3B) was released on January 19, 2026.
Qwen3-Max-Thinking	January 26, 2026	Alibaba Cloud	Unknown	Unknown	Unknown	Proprietary	Proprietary reasoning model with adaptive tool-use, test-time scaling, and iterative self-reflection.[166]
Kimi K2.5	January 27, 2026	Moonshot AI	1040	15T tokens		Modified MIT License	Multimodal MoE with 32B active parameters, derived from Kimi K2.[167] Can use "Agent Swarm" technology to coordinate up to 100 parallel sub-agents.[168][169]
Claude Opus 4.6	February 5, 2026	Anthropic	Unknown	Unknown	Unknown	Proprietary
GPT-5.3-Codex	February 5, 2026	OpenAI	Unknown	Unknown	Unknown	Proprietary
GLM-5	February 12, 2026	Zhipu AI	754			MIT	Specialized for agentic engineering and long-horizon tasks. Integrates DeepSeek Sparse Attention (DSA) for 200K context.
Param-2	February 17, 2026	BharatGen	17	~22T tokens	Unknown	Unknown	Mixture-of-experts model, successor of Param-1; many more Indic languages are supported. Trained on H100 GPUs for 24 days.[170]
Sarvam-1[171]	February 18, 2026[h]	Sarvam AI	105	~12T Tokens	Unknown	Apache 2.0	India's first independently-trained foundation model; has 105B and 30B versions.[173] Based on mixture-of-experts model, using only 10.3B active parameters at a time.[174] Superior in Indic languages.[compared to?]

Close

• List of chatbots
• List of language model benchmarks