Alternatives to OPT
Compare OPT alternatives for your business or organization using the curated list below. SourceForge ranks the best alternatives to OPT in 2026. Compare features, ratings, user reviews, pricing, and more from OPT competitors and alternatives in order to make an informed decision for your business.
-
1
Galactica
Meta
Information overload is a major obstacle to scientific progress. The explosive growth in scientific literature and data has made it ever harder to discover useful insights in a large mass of information. Today scientific knowledge is accessed through search engines, but they are unable to organize scientific knowledge alone. Galactica is a large language model that can store, combine and reason about scientific knowledge. We train on a large scientific corpus of papers, reference material, knowledge bases and many other sources. We outperform existing models on a range of scientific tasks. On technical knowledge probes such as LaTeX equations, Galactica outperforms the latest GPT-3 by 68.2% versus 49.0%. Galactica also performs well on reasoning, outperforming Chinchilla on mathematical MMLU by 41.3% to 35.7%, and PaLM 540B on MATH with a score of 20.4% versus 8.8%. -
2
T5
Google
With T5, we propose reframing all NLP tasks into a unified text-to-text-format where the input and output are always text strings, in contrast to BERT-style models that can only output either a class label or a span of the input. Our text-to-text framework allows us to use the same model, loss function, and hyperparameters on any NLP task, including machine translation, document summarization, question answering, and classification tasks (e.g., sentiment analysis). We can even apply T5 to regression tasks by training it to predict the string representation of a number instead of the number itself. -
3
CodeQwen
Alibaba
CodeQwen is the code version of Qwen, the large language model series developed by the Qwen team, Alibaba Cloud. It is a transformer-based decoder-only language model pre-trained on a large amount of data of codes. Strong code generation capabilities and competitive performance across a series of benchmarks. Supporting long context understanding and generation with the context length of 64K tokens. CodeQwen supports 92 coding languages and provides excellent performance in text-to-SQL, bug fixes, etc. You can just write several lines of code with transformers to chat with CodeQwen. Essentially, we build the tokenizer and the model from pre-trained methods, and we use the generate method to perform chatting with the help of the chat template provided by the tokenizer. We apply the ChatML template for chat models following our previous practice. The model completes the code snippets according to the given prompts, without any additional formatting.Starting Price: Free -
4
PanGu-α
Huawei
PanGu-α is developed under the MindSpore and trained on a cluster of 2048 Ascend 910 AI processors. The training parallelism strategy is implemented based on MindSpore Auto-parallel, which composes five parallelism dimensions to scale the training task to 2048 processors efficiently, including data parallelism, op-level model parallelism, pipeline model parallelism, optimizer model parallelism and rematerialization. To enhance the generalization ability of PanGu-α, we collect 1.1TB high-quality Chinese data from a wide range of domains to pretrain the model. We empirically test the generation ability of PanGu-α in various scenarios including text summarization, question answering, dialogue generation, etc. Moreover, we investigate the effect of model scales on the few-shot performances across a broad range of Chinese NLP tasks. The experimental results demonstrate the superior capabilities of PanGu-α in performing various tasks under few-shot or zero-shot settings. -
5
Falcon-40B
Technology Innovation Institute (TII)
Falcon-40B is a 40B parameters causal decoder-only model built by TII and trained on 1,000B tokens of RefinedWeb enhanced with curated corpora. It is made available under the Apache 2.0 license. Why use Falcon-40B? It is the best open-source model currently available. Falcon-40B outperforms LLaMA, StableLM, RedPajama, MPT, etc. See the OpenLLM Leaderboard. It features an architecture optimized for inference, with FlashAttention and multiquery. It is made available under a permissive Apache 2.0 license allowing for commercial use, without any royalties or restrictions. ⚠️ This is a raw, pretrained model, which should be further finetuned for most usecases. If you are looking for a version better suited to taking generic instructions in a chat format, we recommend taking a look at Falcon-40B-Instruct.Starting Price: Free -
6
Llama 2
Meta
The next generation of our open source large language model. This release includes model weights and starting code for pretrained and fine-tuned Llama language models — ranging from 7B to 70B parameters. Llama 2 pretrained models are trained on 2 trillion tokens, and have double the context length than Llama 1. Its fine-tuned models have been trained on over 1 million human annotations. Llama 2 outperforms other open source language models on many external benchmarks, including reasoning, coding, proficiency, and knowledge tests. Llama 2 was pretrained on publicly available online data sources. The fine-tuned model, Llama-2-chat, leverages publicly available instruction datasets and over 1 million human annotations. We have a broad range of supporters around the world who believe in our open approach to today’s AI — companies that have given early feedback and are excited to build with Llama 2.Starting Price: Free -
7
Qwen-7B
Alibaba
Qwen-7B is the 7B-parameter version of the large language model series, Qwen (abbr. Tongyi Qianwen), proposed by Alibaba Cloud. Qwen-7B is a Transformer-based large language model, which is pretrained on a large volume of data, including web texts, books, codes, etc. Additionally, based on the pretrained Qwen-7B, we release Qwen-7B-Chat, a large-model-based AI assistant, which is trained with alignment techniques. The features of the Qwen-7B series include: Trained with high-quality pretraining data. We have pretrained Qwen-7B on a self-constructed large-scale high-quality dataset of over 2.2 trillion tokens. The dataset includes plain texts and codes, and it covers a wide range of domains, including general domain data and professional domain data. Strong performance. In comparison with the models of the similar model size, we outperform the competitors on a series of benchmark datasets, which evaluates natural language understanding, mathematics, coding, etc. And more.Starting Price: Free -
8
ERNIE 3.0 Titan
Baidu
Pre-trained language models have achieved state-of-the-art results in various Natural Language Processing (NLP) tasks. GPT-3 has shown that scaling up pre-trained language models can further exploit their enormous potential. A unified framework named ERNIE 3.0 was recently proposed for pre-training large-scale knowledge enhanced models and trained a model with 10 billion parameters. ERNIE 3.0 outperformed the state-of-the-art models on various NLP tasks. In order to explore the performance of scaling up ERNIE 3.0, we train a hundred-billion-parameter model called ERNIE 3.0 Titan with up to 260 billion parameters on the PaddlePaddle platform. Furthermore, We design a self-supervised adversarial loss and a controllable language modeling loss to make ERNIE 3.0 Titan generate credible and controllable texts. -
9
Orpheus TTS
Canopy Labs
Canopy Labs has introduced Orpheus, a family of state-of-the-art speech large language models (LLMs) designed for human-level speech generation. These models are built on the Llama-3 architecture and are trained on over 100,000 hours of English speech data, enabling them to produce natural intonation, emotion, and rhythm that surpasses current state-of-the-art closed source models. Orpheus supports zero-shot voice cloning, allowing users to replicate voices without prior fine-tuning, and offers guided emotion and intonation control through simple tags. The models achieve low latency, with approximately 200ms streaming latency for real-time applications, reducible to around 100ms with input streaming. Canopy Labs has released both pre-trained and fine-tuned 3B-parameter models under the permissive Apache 2.0 license, with plans to release smaller models of 1B, 400M, and 150M parameters for use on resource-constrained devices. -
10
GPT-4
OpenAI
GPT-4 (Generative Pre-trained Transformer 4) is a large-scale unsupervised language model, yet to be released by OpenAI. GPT-4 is the successor to GPT-3 and part of the GPT-n series of natural language processing models, and was trained on a dataset of 45TB of text to produce human-like text generation and understanding capabilities. Unlike most other NLP models, GPT-4 does not require additional training data for specific tasks. Instead, it can generate text or answer questions using only its own internally generated context as input. GPT-4 has been shown to be able to perform a wide variety of tasks without any task specific training data such as translation, summarization, question answering, sentiment analysis and more.Starting Price: $0.0200 per 1000 tokens -
11
Olmo 3
Ai2
Olmo 3 is a fully open model family spanning 7 billion and 32 billion parameter variants that delivers not only high-performing base, reasoning, instruction, and reinforcement-learning models, but also exposure of the entire model flow, including raw training data, intermediate checkpoints, training code, long-context support (65,536 token window), and provenance tooling. Starting with the Dolma 3 dataset (≈9 trillion tokens) and its disciplined mix of web text, scientific PDFs, code, and long-form documents, the pre-training, mid-training, and long-context phases shape the base models, which are then post-trained via supervised fine-tuning, direct preference optimisation, and RL with verifiable rewards to yield the Think and Instruct variants. The 32 B Think model is described as the strongest fully open reasoning model to date, competitively close to closed-weight peers in math, code, and complex reasoning.Starting Price: Free -
12
Falcon-7B
Technology Innovation Institute (TII)
Falcon-7B is a 7B parameters causal decoder-only model built by TII and trained on 1,500B tokens of RefinedWeb enhanced with curated corpora. It is made available under the Apache 2.0 license. Why use Falcon-7B? It outperforms comparable open-source models (e.g., MPT-7B, StableLM, RedPajama etc.), thanks to being trained on 1,500B tokens of RefinedWeb enhanced with curated corpora. See the OpenLLM Leaderboard. It features an architecture optimized for inference, with FlashAttention and multiquery. It is made available under a permissive Apache 2.0 license allowing for commercial use, without any royalties or restrictions.Starting Price: Free -
13
Azure OpenAI Service
Microsoft
Apply advanced coding and language models to a variety of use cases. Leverage large-scale, generative AI models with deep understandings of language and code to enable new reasoning and comprehension capabilities for building cutting-edge applications. Apply these coding and language models to a variety of use cases, such as writing assistance, code generation, and reasoning over data. Detect and mitigate harmful use with built-in responsible AI and access enterprise-grade Azure security. Gain access to generative models that have been pretrained with trillions of words. Apply them to new scenarios including language, code, reasoning, inferencing, and comprehension. Customize generative models with labeled data for your specific scenario using a simple REST API. Fine-tune your model's hyperparameters to increase accuracy of outputs. Use the few-shot learning capability to provide the API with examples and achieve more relevant results.Starting Price: $0.0004 per 1000 tokens -
14
ALBERT
Google
ALBERT is a self-supervised Transformer model that was pretrained on a large corpus of English data. This means it does not require manual labelling, and instead uses an automated process to generate inputs and labels from raw texts. It is trained with two distinct objectives in mind. The first is Masked Language Modeling (MLM), which randomly masks 15% of words in the input sentence and requires the model to predict them. This technique differs from RNNs and autoregressive models like GPT as it allows the model to learn bidirectional sentence representations. The second objective is Sentence Ordering Prediction (SOP), which entails predicting the ordering of two consecutive segments of text during pretraining. -
15
Baichuan-13B
Baichuan Intelligent Technology
Baichuan-13B is an open source and commercially available large-scale language model containing 13 billion parameters developed by Baichuan Intelligent following Baichuan -7B . It has achieved the best results of the same size on authoritative Chinese and English benchmarks. This release contains two versions of pre-training ( Baichuan-13B-Base ) and alignment ( Baichuan-13B-Chat ). Larger size, more data : Baichuan-13B further expands the number of parameters to 13 billion on the basis of Baichuan -7B , and trains 1.4 trillion tokens on high-quality corpus, which is 40% more than LLaMA-13B. It is currently open source The model with the largest amount of training data in the 13B size. Support Chinese and English bilingual, use ALiBi position code, context window length is 4096.Starting Price: Free -
16
BERT
Google
BERT is a large language model and a method of pre-training language representations. Pre-training refers to how BERT is first trained on a large source of text, such as Wikipedia. You can then apply the training results to other Natural Language Processing (NLP) tasks, such as question answering and sentiment analysis. With BERT and AI Platform Training, you can train a variety of NLP models in about 30 minutes.Starting Price: Free -
17
Yi-Large
01.AI
Yi-Large is a proprietary large language model developed by 01.AI, offering a 32k context length with both input and output costs at $2 per million tokens. It stands out with its advanced capabilities in natural language processing, common-sense reasoning, and multilingual support, performing on par with leading models like GPT-4 and Claude3 in various benchmarks. Yi-Large is designed for tasks requiring complex inference, prediction, and language understanding, making it suitable for applications like knowledge search, data classification, and creating human-like chatbots. Its architecture is based on a decoder-only transformer with enhancements such as pre-normalization and Group Query Attention, and it has been trained on a vast, high-quality multilingual dataset. This model's versatility and cost-efficiency make it a strong contender in the AI market, particularly for enterprises aiming to deploy AI solutions globally.Starting Price: $0.19 per 1M input token -
18
DeepSeek-V2
DeepSeek
DeepSeek-V2 is a state-of-the-art Mixture-of-Experts (MoE) language model introduced by DeepSeek-AI, characterized by its economical training and efficient inference capabilities. With a total of 236 billion parameters, of which only 21 billion are active per token, it supports a context length of up to 128K tokens. DeepSeek-V2 employs innovative architectures like Multi-head Latent Attention (MLA) for efficient inference by compressing the Key-Value (KV) cache and DeepSeekMoE for cost-effective training through sparse computation. This model significantly outperforms its predecessor, DeepSeek 67B, by saving 42.5% in training costs, reducing the KV cache by 93.3%, and enhancing generation throughput by 5.76 times. Pretrained on an 8.1 trillion token corpus, DeepSeek-V2 excels in language understanding, coding, and reasoning tasks, making it a top-tier performer among open-source models.Starting Price: Free -
19
TinyLlama
TinyLlama
The TinyLlama project aims to pretrain a 1.1B Llama model on 3 trillion tokens. With some proper optimization, we can achieve this within a span of "just" 90 days using 16 A100-40G GPUs. We adopted exactly the same architecture and tokenizer as Llama 2. This means TinyLlama can be plugged and played in many open-source projects built upon Llama. Besides, TinyLlama is compact with only 1.1B parameters. This compactness allows it to cater to a multitude of applications demanding a restricted computation and memory footprint.Starting Price: Free -
20
Hippocratic AI
Hippocratic AI
Hippocratic AI is the new state of the art (SOTA) model, outperforming GPT-4 on 105 of 114 healthcare exams and certifications. Hippocratic AI has outperformed GPT-4 on 105 out of 114 tests and certifications, outperformed by a margin of five percent or more on 74 of the certifications, and outperformed by a margin of ten percent or more on 43 of the certifications. Most language models pre-train on the common crawl of the Internet, which may include incorrect and misleading information. Unlike these LLMs, Hippocratic AI is investing heavily in legally acquiring evidence-based healthcare content. We’re conducting a unique Reinforcement Learning with Human Feedback process using healthcare professionals to train and validate the model’s readiness for deployment. We call this RLHF-HP. Hippocratic AI will not release the model until a large number of these licensed professionals deem it safe. -
21
Granite Code
IBM
We introduce the Granite series of decoder-only code models for code generative tasks (e.g., fixing bugs, explaining code, documenting code), trained with code written in 116 programming languages. A comprehensive evaluation of the Granite Code model family on diverse tasks demonstrates that our models consistently reach state-of-the-art performance among available open source code LLMs. The key advantages of Granite Code models include: All-rounder Code LLM: Granite Code models achieve competitive or state-of-the-art performance on different kinds of code-related tasks, including code generation, explanation, fixing, editing, translation, and more. Demonstrating their ability to solve diverse coding tasks. Trustworthy Enterprise-Grade LLM: All our models are trained on license-permissible data collected following IBM's AI Ethics principles and guided by IBM’s Corporate Legal team for trustworthy enterprise usage.Starting Price: Free -
22
Megatron-Turing
NVIDIA
Megatron-Turing Natural Language Generation model (MT-NLG), is the largest and the most powerful monolithic transformer English language model with 530 billion parameters. This 105-layer, transformer-based MT-NLG improves upon the prior state-of-the-art models in zero-, one-, and few-shot settings. It demonstrates unmatched accuracy in a broad set of natural language tasks such as, Completion prediction, Reading comprehension, Commonsense reasoning, Natural language inferences, Word sense disambiguation, etc. With the intent of accelerating research on the largest English language model till date and enabling customers to experiment, employ and apply such a large language model on downstream language tasks - NVIDIA is pleased to announce an Early Access program for its managed API service to MT-NLG mode. -
23
PanGu-Σ
Huawei
Significant advancements in the field of natural language processing, understanding, and generation have been achieved through the expansion of large language models. This study introduces a system which utilizes Ascend 910 AI processors and the MindSpore framework to train a language model with over a trillion parameters, specifically 1.085T, named PanGu-{\Sigma}. This model, which builds upon the foundation laid by PanGu-{\alpha}, takes the traditionally dense Transformer model and transforms it into a sparse one using a concept known as Random Routed Experts (RRE). The model was efficiently trained on a dataset of 329 billion tokens using a technique called Expert Computation and Storage Separation (ECSS), leading to a 6.3-fold increase in training throughput via heterogeneous computing. Experimentation indicates that PanGu-{\Sigma} sets a new standard in zero-shot learning for various downstream Chinese NLP tasks. -
24
Cerebras-GPT
Cerebras
State-of-the-art language models are extremely challenging to train; they require huge compute budgets, complex distributed compute techniques and deep ML expertise. As a result, few organizations train large language models (LLMs) from scratch. And increasingly those that have the resources and expertise are not open sourcing the results, marking a significant change from even a few months back. At Cerebras, we believe in fostering open access to the most advanced models. With this in mind, we are proud to announce the release to the open source community of Cerebras-GPT, a family of seven GPT models ranging from 111 million to 13 billion parameters. Trained using the Chinchilla formula, these models provide the highest accuracy for a given compute budget. Cerebras-GPT has faster training times, lower training costs, and consumes less energy than any publicly available model to date.Starting Price: Free -
25
Olmo 2
Ai2
Olmo 2 is a family of fully open language models developed by the Allen Institute for AI (AI2), designed to provide researchers and developers with transparent access to training data, open-source code, reproducible training recipes, and comprehensive evaluations. These models are trained on up to 5 trillion tokens and are competitive with leading open-weight models like Llama 3.1 on English academic benchmarks. Olmo 2 emphasizes training stability, implementing techniques to prevent loss spikes during long training runs, and utilizes staged training interventions during late pretraining to address capability deficiencies. The models incorporate state-of-the-art post-training methodologies from AI2's Tülu 3, resulting in the creation of Olmo 2-Instruct models. An actionable evaluation framework, the Open Language Modeling Evaluation System (OLMES), was established to guide improvements through development stages, consisting of 20 evaluation benchmarks assessing core capabilities. -
26
Mistral NeMo
Mistral AI
Mistral NeMo, our new best small model. A state-of-the-art 12B model with 128k context length, and released under the Apache 2.0 license. Mistral NeMo is a 12B model built in collaboration with NVIDIA. Mistral NeMo offers a large context window of up to 128k tokens. Its reasoning, world knowledge, and coding accuracy are state-of-the-art in its size category. As it relies on standard architecture, Mistral NeMo is easy to use and a drop-in replacement in any system using Mistral 7B. We have released pre-trained base and instruction-tuned checkpoints under the Apache 2.0 license to promote adoption for researchers and enterprises. Mistral NeMo was trained with quantization awareness, enabling FP8 inference without any performance loss. The model is designed for global, multilingual applications. It is trained on function calling and has a large context window. Compared to Mistral 7B, it is much better at following precise instructions, reasoning, and handling multi-turn conversations.Starting Price: Free -
27
Giga ML
Giga ML
We just launched X1 large series of Models. Giga ML's most powerful model is available for pre-training and fine-tuning with on-prem deployment. Since we are Open AI compatible, your existing integrations with long chain, llama-index, and all others work seamlessly. You can continue pre-training of LLM's with domain-specific data books or docs or company docs. The world of large language models (LLMs) rapidly expanding, offering unprecedented opportunities for natural language processing across various domains. However, some critical challenges have remained unaddressed. At Giga ML, we proudly introduce the X1 Large 32k model, a pioneering on-premise LLM solution that addresses these critical issues. -
28
GLM-5
Zhipu AI
GLM-5 is Z.ai’s latest large language model built for complex systems engineering and long-horizon agentic tasks. It scales significantly beyond GLM-4.5, increasing total parameters and training data while integrating DeepSeek Sparse Attention to reduce deployment costs without sacrificing long-context capacity. The model combines enhanced pre-training with a new asynchronous reinforcement learning infrastructure called slime, improving training efficiency and post-training refinement. GLM-5 achieves best-in-class performance among open-source models across reasoning, coding, and agent benchmarks, narrowing the gap with leading frontier models. It ranks highly on evaluations such as Vending Bench 2, demonstrating strong long-term planning and operational capabilities. The model is open-sourced under the MIT License.Starting Price: Free -
29
CodeGemma
Google
CodeGemma is a collection of powerful, lightweight models that can perform a variety of coding tasks like fill-in-the-middle code completion, code generation, natural language understanding, mathematical reasoning, and instruction following. CodeGemma has 3 model variants, a 7B pre-trained variant that specializes in code completion and generation from code prefixes and/or suffixes, a 7B instruction-tuned variant for natural language-to-code chat and instruction following; and a state-of-the-art 2B pre-trained variant that provides up to 2x faster code completion. Complete lines, and functions, and even generate entire blocks of code, whether you're working locally or using Google Cloud resources. Trained on 500 billion tokens of primarily English language data from web documents, mathematics, and code, CodeGemma models generate code that's not only more syntactically correct but also semantically meaningful, reducing errors and debugging time. -
30
InstructGPT
OpenAI
InstructGPT is an open-source framework for training language models to generate natural language instructions from visual input. It uses a generative pre-trained transformer (GPT) model and the state-of-the-art object detector, Mask R-CNN, to detect objects in images and generate natural language sentences that describe the image. InstructGPT is designed to be effective across domains such as robotics, gaming and education; it can assist robots in navigating complex tasks with natural language instructions, or help students learn by providing descriptive explanations of processes or events.Starting Price: $0.0200 per 1000 tokens -
31
Yi-Lightning
Yi-Lightning
Yi-Lightning, developed by 01.AI under the leadership of Kai-Fu Lee, represents the latest advancement in large language models with a focus on high performance and cost-efficiency. It boasts a maximum context length of 16K tokens and is priced at $0.14 per million tokens for both input and output, making it remarkably competitive. Yi-Lightning leverages an enhanced Mixture-of-Experts (MoE) architecture, incorporating fine-grained expert segmentation and advanced routing strategies, which contribute to its efficiency in training and inference. This model has excelled in various domains, achieving top rankings in categories like Chinese, math, coding, and hard prompts on the chatbot arena, where it secured the 6th position overall and 9th in style control. Its development included comprehensive pre-training, supervised fine-tuning, and reinforcement learning from human feedback, ensuring both performance and safety, with optimizations in memory usage and inference speed. -
32
Llama 3.2
Meta
The open-source AI model you can fine-tune, distill and deploy anywhere is now available in more versions. Choose from 1B, 3B, 11B or 90B, or continue building with Llama 3.1. Llama 3.2 is a collection of large language models (LLMs) pretrained and fine-tuned in 1B and 3B sizes that are multilingual text only, and 11B and 90B sizes that take both text and image inputs and output text. Develop highly performative and efficient applications from our latest release. Use our 1B or 3B models for on device applications such as summarizing a discussion from your phone or calling on-device tools like calendar. Use our 11B or 90B models for image use cases such as transforming an existing image into something new or getting more information from an image of your surroundings.Starting Price: Free -
33
Teuken 7B
OpenGPT-X
Teuken-7B is a multilingual, open source language model developed under the OpenGPT-X initiative, specifically designed to cater to Europe's diverse linguistic landscape. It has been trained on a dataset comprising over 50% non-English texts, encompassing all 24 official languages of the European Union, ensuring robust performance across these languages. A key innovation in Teuken-7B is its custom multilingual tokenizer, optimized for European languages, which enhances training efficiency and reduces inference costs compared to standard monolingual tokenizers. The model is available in two versions, Teuken-7B-Base, the foundational pre-trained model, and Teuken-7B-Instruct, which has undergone instruction tuning for improved performance in following user prompts. Both versions are accessible on Hugging Face, promoting transparency and collaboration within the AI community. The development of Teuken-7B underscores a commitment to creating AI models that reflect Europe's diversity.Starting Price: Free -
34
Mu
Microsoft
Mu is a 330-million-parameter encoder–decoder language model designed to power the agent in Windows settings by mapping natural-language queries to Settings function calls, running fully on-device via NPUs at over 100 tokens per second while maintaining high accuracy. Drawing on Phi Silica optimizations, Mu’s encoder–decoder architecture reuses a fixed-length latent representation to cut computation and memory overhead, yielding 47 percent lower first-token latency and 4.7× higher decoding speed on Qualcomm Hexagon NPUs compared to similar decoder-only models. Hardware-aware tuning, including a 2/3–1/3 encoder–decoder parameter split, weight sharing between input and output embeddings, Dual LayerNorm, rotary positional embeddings, and grouped-query attention, enables fast inference at over 200 tokens per second on devices like Surface Laptop 7 and sub-500 ms response times for settings queries. -
35
Kimi K2
Moonshot AI
Kimi K2 is a state-of-the-art open source large language model series built on a mixture-of-experts (MoE) architecture, featuring 1 trillion total parameters and 32 billion activated parameters for task-specific efficiency. Trained with the Muon optimizer on over 15.5 trillion tokens and stabilized by MuonClip’s attention-logit clamping, it delivers exceptional performance in frontier knowledge, reasoning, mathematics, coding, and general agentic workflows. Moonshot AI provides two variants, Kimi-K2-Base for research-level fine-tuning and Kimi-K2-Instruct pre-trained for immediate chat and tool-driven interactions, enabling both custom development and drop-in agentic capabilities. Benchmarks show it outperforms leading open source peers and rivals top proprietary models in coding tasks and complex task breakdowns, while its 128 K-token context length, tool-calling API compatibility, and support for industry-standard inference engines.Starting Price: Free -
36
mT5
Google
Multilingual T5 (mT5) is a massively multilingual pretrained text-to-text transformer model, trained following a similar recipe as T5. This repo can be used to reproduce the experiments in the mT5 paper. mT5 is pretrained on the mC4 corpus, covering 101 languages: Afrikaans, Albanian, Amharic, Arabic, Armenian, Azerbaijani, Basque, Belarusian, Bengali, Bulgarian, Burmese, Catalan, Cebuano, Chichewa, Chinese, Corsican, Czech, Danish, Dutch, English, Esperanto, Estonian, Filipino, Finnish, French, Galician, Georgian, German, Greek, Gujarati, Haitian Creole, Hausa, Hawaiian, Hebrew, Hindi, Hmong, Hungarian, Icelandic, Igbo, Indonesian, Irish, Italian, Japanese, Javanese, Kannada, Kazakh, Khmer, Korean, Kurdish, Kyrgyz, Lao, Latin, Latvian, Lithuanian, Luxembourgish, Macedonian, Malagasy, Malay, Malayalam, Maltese, Maori, Marathi, Mongolian, Nepali, Norwegian, Pashto, Persian, Polish, Portuguese, Punjabi, Romanian, Russian, Samoan, Scottish Gaelic, Serbian, Shona, Sindhi, and more.Starting Price: Free -
37
Stable LM
Stability AI
Stable LM: Stability AI Language Models. The release of Stable LM builds on our experience in open-sourcing earlier language models with EleutherAI, a nonprofit research hub. These language models include GPT-J, GPT-NeoX, and the Pythia suite, which were trained on The Pile open-source dataset. Many recent open-source language models continue to build on these efforts, including Cerebras-GPT and Dolly-2. Stable LM is trained on a new experimental dataset built on The Pile, but three times larger with 1.5 trillion tokens of content. We will release details on the dataset in due course. The richness of this dataset gives Stable LM surprisingly high performance in conversational and coding tasks, despite its small size of 3 to 7 billion parameters (by comparison, GPT-3 has 175 billion parameters). Stable LM 3B is a compact language model designed to operate on portable digital devices like handhelds and laptops, and we’re excited about its capabilities and portability.Starting Price: Free -
38
MAI-1-preview
Microsoft
MAI-1 Preview is Microsoft AI’s first end-to-end trained foundation model, built entirely in-house as a mixture-of-experts architecture. Pre-trained and post-trained on approximately 15,000 NVIDIA H100 GPUs, it is designed to follow instructions and generate helpful, responsive text for everyday user queries, representing a prototype of future Copilot capabilities. Now available for public testing on LMArena, MAI-1 Preview delivers an early glimpse into the platform’s trajectory, with plans to roll out select text-based applications within Copilot over the coming weeks to gather user feedback and refine performance. Microsoft reinforces that it will continue combining its own models, partner models, and developments from the open-source community to flexibly power experiences across millions of unique interactions each day. -
39
Sky-T1
NovaSky
Sky-T1-32B-Preview is an open source reasoning model developed by the NovaSky team at UC Berkeley's Sky Computing Lab. It matches the performance of proprietary models like o1-preview on reasoning and coding benchmarks, yet was trained for under $450, showcasing the feasibility of cost-effective, high-level reasoning capabilities. The model was fine-tuned from Qwen2.5-32B-Instruct using a curated dataset of 17,000 examples across diverse domains, including math and coding. The training was completed in 19 hours on eight H100 GPUs with DeepSpeed Zero-3 offloading. All aspects of the project, including data, code, and model weights, are fully open-source, empowering the academic and open-source communities to replicate and enhance the model's performance.Starting Price: Free -
40
RoBERTa
Meta
RoBERTa builds on BERT’s language masking strategy, wherein the system learns to predict intentionally hidden sections of text within otherwise unannotated language examples. RoBERTa, which was implemented in PyTorch, modifies key hyperparameters in BERT, including removing BERT’s next-sentence pretraining objective, and training with much larger mini-batches and learning rates. This allows RoBERTa to improve on the masked language modeling objective compared with BERT and leads to better downstream task performance. We also explore training RoBERTa on an order of magnitude more data than BERT, for a longer amount of time. We used existing unannotated NLP datasets as well as CC-News, a novel set drawn from public news articles.Starting Price: Free -
41
Qwen3
Alibaba
Qwen3, the latest iteration of the Qwen family of large language models, introduces groundbreaking features that enhance performance across coding, math, and general capabilities. With models like the Qwen3-235B-A22B and Qwen3-30B-A3B, Qwen3 achieves impressive results compared to top-tier models, thanks to its hybrid thinking modes that allow users to control the balance between deep reasoning and quick responses. The platform supports 119 languages and dialects, making it an ideal choice for global applications. Its pre-training process, which uses 36 trillion tokens, enables robust performance, and advanced reinforcement learning (RL) techniques continue to refine its capabilities. Available on platforms like Hugging Face and ModelScope, Qwen3 offers a powerful tool for developers and researchers working in diverse fields.Starting Price: Free -
42
Jamba
AI21 Labs
Jamba is the most powerful & efficient long context model, open for builders and built for the enterprise. Jamba's latency outperforms all leading models of comparable sizes. Jamba's 256k context window is the longest openly available. Jamba's Mamba-Transformer MoE architecture is designed for cost & efficiency gains. Jamba offers key features of OOTB including function calls, JSON mode output, document objects, and citation mode. Jamba 1.5 models maintain high performance across the full length of their context window. Jamba 1.5 models achieve top scores across common quality benchmarks. Secure deployment that suits your enterprise. Seamlessly start using Jamba on our production-grade SaaS platform. The Jamba model family is available for deployment across our strategic partners. We offer VPC & on-prem deployments for enterprises that require custom solutions. For enterprises that have unique, bespoke requirements, we offer hands-on management, continuous pre-training, etc. -
43
DeepSeek-V3.1-Terminus
DeepSeek
DeepSeek has released DeepSeek-V3.1-Terminus, which enhances the V3.1 architecture by incorporating user feedback to improve output stability, consistency, and agent performance. It notably reduces instances of mixed Chinese/English character output and unintended garbled characters, resulting in cleaner, more consistent language generation. The update upgrades both the code agent and search agent subsystems to yield stronger, more reliable performance across benchmarks. DeepSeek-V3.1-Terminus is also available as an open source model, and its weights are published on Hugging Face. The model structure remains the same as DeepSeek-V3, ensuring compatibility with existing deployment methods, with updated inference demos provided for community use. While trained at a scale of 685B parameters, the model includes FP8, BF16, and F32 tensor formats, offering flexibility across environments.Starting Price: Free -
44
Qwen3-Max
Alibaba
Qwen3-Max is Alibaba’s latest trillion-parameter large language model, designed to push performance in agentic tasks, coding, reasoning, and long-context processing. It is built atop the Qwen3 family and benefits from the architectural, training, and inference advances introduced there; mixing thinker and non-thinker modes, a “thinking budget” mechanism, and support for dynamic mode switching based on complexity. The model reportedly processes extremely long inputs (hundreds of thousands of tokens), supports tool invocation, and exhibits strong performance on benchmarks in coding, multi-step reasoning, and agent benchmarks (e.g., Tau2-Bench). While its initial variant emphasizes instruction following (non-thinking mode), Alibaba plans to bring reasoning capabilities online to enable autonomous agent behavior. Qwen3-Max inherits multilingual support and extensive pretraining on trillions of tokens, and it is delivered via API interfaces compatible with OpenAI-style functions.Starting Price: Free -
45
Ministral 3
Mistral AI
Mistral 3 is the latest generation of open-weight AI models from Mistral AI, offering a full family of models, from small, edge-optimized versions to a flagship, large-scale multimodal model. The lineup includes three compact “Ministral 3” models (3B, 8B, and 14B parameters) designed for efficiency and deployment on constrained hardware (even laptops, drones, or edge devices), plus the powerful “Mistral Large 3,” a sparse mixture-of-experts model with 675 billion total parameters (41 billion active). The models support multimodal and multilingual tasks, not only text, but also image understanding, and have demonstrated best-in-class performance on general prompts, multilingual conversations, and multimodal inputs. The base and instruction-fine-tuned versions are released under the Apache 2.0 license, enabling broad customization and integration in enterprise and open source projects.Starting Price: Free -
46
Grok 3 DeepSearch is an advanced model and research agent designed to improve reasoning and problem-solving abilities in AI, with a strong focus on deep search and iterative reasoning. Unlike traditional models that rely solely on pre-trained knowledge, Grok 3 DeepSearch can explore multiple avenues, test hypotheses, and correct errors in real-time by analyzing vast amounts of information and engaging in chain-of-thought processes. It is designed for tasks that require critical thinking, such as complex mathematical problems, coding challenges, and intricate academic inquiries. Grok 3 DeepSearch is a cutting-edge AI tool capable of providing accurate and thorough solutions by using its unique deep search capabilities, making it ideal for both STEM and creative fields.Starting Price: $30/month
-
47
Phi-2
Microsoft
We are now releasing Phi-2, a 2.7 billion-parameter language model that demonstrates outstanding reasoning and language understanding capabilities, showcasing state-of-the-art performance among base language models with less than 13 billion parameters. On complex benchmarks Phi-2 matches or outperforms models up to 25x larger, thanks to new innovations in model scaling and training data curation. With its compact size, Phi-2 is an ideal playground for researchers, including for exploration around mechanistic interpretability, safety improvements, or fine-tuning experimentation on a variety of tasks. We have made Phi-2 available in the Azure AI Studio model catalog to foster research and development on language models. -
48
Sparrow
DeepMind
Sparrow is a research model and proof of concept, designed with the goal of training dialogue agents to be more helpful, correct, and harmless. By learning these qualities in a general dialogue setting, Sparrow advances our understanding of how we can train agents to be safer and more useful – and ultimately, to help build safer and more useful artificial general intelligence (AGI). Sparrow is not yet available for public use. Training a conversational AI is an especially challenging problem because it’s difficult to pinpoint what makes a dialogue successful. To address this problem, we turn to a form of reinforcement learning (RL) based on people's feedback, using the study participants’ preference feedback to train a model of how useful an answer is. To get this data, we show our participants multiple model answers to the same question and ask them which answer they like the most. -
49
MPT-7B
MosaicML
Introducing MPT-7B, the latest entry in our MosaicML Foundation Series. MPT-7B is a transformer trained from scratch on 1T tokens of text and code. It is open source, available for commercial use, and matches the quality of LLaMA-7B. MPT-7B was trained on the MosaicML platform in 9.5 days with zero human intervention at a cost of ~$200k. Now you can train, finetune, and deploy your own private MPT models, either starting from one of our checkpoints or training from scratch. For inspiration, we are also releasing three finetuned models in addition to the base MPT-7B: MPT-7B-Instruct, MPT-7B-Chat, and MPT-7B-StoryWriter-65k+, the last of which uses a context length of 65k tokens!Starting Price: Free -
50
MiniMax M1
MiniMax
MiniMax‑M1 is a large‑scale hybrid‑attention reasoning model released by MiniMax AI under the Apache 2.0 license. It supports an unprecedented 1 million‑token context window and up to 80,000-token outputs, enabling extended reasoning across long documents. Trained using large‑scale reinforcement learning with a novel CISPO algorithm, MiniMax‑M1 completed full training on 512 H800 GPUs in about three weeks. It achieves state‑of‑the‑art performance on benchmarks in mathematics, coding, software engineering, tool usage, and long‑context understanding, matching or outperforming leading models. Two model variants are available (40K and 80K thinking budgets), with weights and deployment scripts provided via GitHub and Hugging Face.
