DeepSeek-R1-0528 is a reasoning model powered by reinforcement learning (RL) that addresses the issues of repetition and readability. Prior to RL, DeepSeek-R1 incorporated cold-start data to further optimize its reasoning performance. It achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks, and through carefully designed training methods, it has enhanced overall effectiveness...

DeepSeek-R1 is a reasoning model powered by reinforcement learning (RL) that addresses the issues of repetition and readability. Prior to RL, DeepSeek-R1 incorporated cold-start data to further optimize its reasoning performance. It achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks, and through carefully designed training methods, it has enhanced overall effectiveness...

DeepSeek-R1-Distill-Llama-70B is a distilled model based on Llama-3.3-70B-Instruct. As part of the DeepSeek-R1 series, it was fine-tuned using samples generated by DeepSeek-R1 and demonstrates excellent performance across mathematics, programming, and reasoning tasks. The model achieved impressive results in various benchmarks including AIME 2024, MATH-500, and GPQA Diamond, showcasing its strong reasoning capabilities...

DeepSeek-R1-Distill-Llama-8B is a distilled model based on Llama-3.1-8B. The model was fine-tuned using samples generated by DeepSeek-R1 and demonstrates strong reasoning capabilities. It achieved notable results across various benchmarks, including 89.1% accuracy on MATH-500, 50.4% pass rate on AIME 2024, and a rating of 1205 on CodeForces, showing impressive mathematical and programming abilities for an 8B-scale model...

DeepSeek-R1-Distill-Qwen-1.5B is a distilled model based on Qwen2.5-Math-1.5B. The model was fine-tuned using 800k curated samples generated by DeepSeek-R1 and demonstrates decent performance across various benchmarks. As a lightweight model, it achieved 83.9% accuracy on MATH-500, 28.9% pass rate on AIME 2024, and a rating of 954 on CodeForces, showing reasoning capabilities beyond its parameter scale...

DeepSeek-R1-Distill-Qwen-14B is a distilled model based on Qwen2.5-14B. The model was fine-tuned using 800k curated samples generated by DeepSeek-R1 and demonstrates strong reasoning capabilities. It achieved impressive results across various benchmarks, including 93.9% accuracy on MATH-500, 69.7% pass rate on AIME 2024, and a rating of 1481 on CodeForces, showcasing its powerful abilities in mathematics and programming tasks...

DeepSeek-R1-Distill-Qwen-32B is a distilled model based on Qwen2.5-32B. The model was fine-tuned using 800k curated samples generated by DeepSeek-R1 and demonstrates exceptional performance across mathematics, programming, and reasoning tasks. It achieved impressive results in various benchmarks including AIME 2024, MATH-500, and GPQA Diamond, with a notable 94.3% accuracy on MATH-500, showcasing its strong mathematical reasoning capabilities...

DeepSeek-R1-Distill-Qwen-7B is a distilled model based on Qwen2.5-Math-7B. The model was fine-tuned using 800k curated samples generated by DeepSeek-R1 and demonstrates strong reasoning capabilities. It achieved impressive results across various benchmarks, including 92.8% accuracy on MATH-500, 55.5% pass rate on AIME 2024, and a rating of 1189 on CodeForces, showing remarkable mathematical and programming abilities for a 7B-scale model...

The new version of DeepSeek-V3 (DeepSeek-V3-0324) utilizes the same base model as the previous DeepSeek-V3-1226, with improvements made only to the post-training methods. The new V3 model incorporates reinforcement learning techniques from the training process of the DeepSeek-R1 model, significantly enhancing its performance on reasoning tasks. It has achieved scores surpassing GPT-4.5 on evaluation sets related to mathematics and coding. Additionally, the model has seen notable improvements in tool invocation, role-playing, and casual conversation capabilities....

DeepSeek-V3.1 is a hybrid large language model released by DeepSeek AI, featuring significant upgrades over its predecessor. A key innovation is the integration of both a 'Thinking Mode' for deliberative, chain-of-thought reasoning and a 'Non-thinking Mode' for direct responses, which can be switched via the chat template to suit various tasks. The model's capabilities in tool use and agent tasks have been substantially improved through post-training optimization, enabling better support for external search tools and complex multi-step instructions. DeepSeek-V3.1 is post-trained on top of the DeepSeek-V3.1-Base model, which underwent a two-phase long-context extension with a vastly expanded dataset, enhancing its ability to process long documents and codebases. As an open-source model, DeepSeek-V3.1 demonstrates performance comparable to leading closed-source models on various benchmarks, particularly in coding, math, and reasoning, while its Mixture-of-Experts (MoE) architecture maintains a massive parameter count while reducing inference costs...

DeepSeek-VL2 is a mixed-expert (MoE) vision-language model developed based on DeepSeekMoE-27B, employing a sparse-activated MoE architecture to achieve superior performance with only 4.5B active parameters. The model excels in various tasks including visual question answering, optical character recognition, document/table/chart understanding, and visual grounding. Compared to existing open-source dense models and MoE-based models, it demonstrates competitive or state-of-the-art performance using the same or fewer active parameters....

ERNIE-4.5-300B-A47B is a large language model developed by Baidu based on a Mixture-of-Experts (MoE) architecture. The model has a total of 300 billion parameters, but only activates 47 billion parameters per token during inference, thus balancing powerful performance with computational efficiency. As one of the core models in the ERNIE 4.5 series, it is trained on the PaddlePaddle deep learning framework and demonstrates outstanding capabilities in tasks such as text understanding, generation, reasoning, and coding. The model utilizes an innovative multimodal heterogeneous MoE pre-training method, which effectively enhances its overall abilities through joint training on text and visual modalities, showing prominent results in instruction following and world knowledge memorization. Baidu has open-sourced this model along with others in the series to promote the research and application of AI technology...

GLM-4-32B-0414 is a new generation model in the GLM family with 32 billion parameters. Its performance is comparable to OpenAI's GPT series and DeepSeek's V3/R1 series, and it supports very user-friendly local deployment features. GLM-4-32B-Base-0414 was pre-trained on 15T of high-quality data, including a large amount of reasoning-type synthetic data, laying the foundation for subsequent reinforcement learning extensions. In the post-training stage, in addition to human preference alignment for dialogue scenarios, the team enhanced the model's performance in instruction following, engineering code, and function calling using techniques such as rejection sampling and reinforcement learning, strengthening the atomic capabilities required for agent tasks. GLM-4-32B-0414 achieves good results in areas such as engineering code, Artifact generation, function calling, search-based Q&A, and report generation. On several benchmarks, its performance approaches or even exceeds that of larger models like GPT-4o and DeepSeek-V3-0324 (671B)...

GLM-4-9B-0414 is a small-sized model in the GLM series with 9 billion parameters. This model inherits the technical characteristics of the GLM-4-32B series but offers a more lightweight deployment option. Despite its smaller scale, GLM-4-9B-0414 still demonstrates excellent capabilities in code generation, web design, SVG graphics generation, and search-based writing tasks. The model also supports function calling features, allowing it to invoke external tools to extend its range of capabilities. The model shows a good balance between efficiency and effectiveness in resource-constrained scenarios, providing a powerful option for users who need to deploy AI models under limited computational resources. Like other models in the same series, GLM-4-9B-0414 also demonstrates competitive performance in various benchmark tests...

GLM-4.1V-9B-Thinking is an open-source Vision-Language Model (VLM) jointly released by Zhipu AI and Tsinghua University's KEG lab, designed to advance general-purpose multimodal reasoning. Built upon the GLM-4-9B-0414 foundation model, it introduces a 'thinking paradigm' and leverages Reinforcement Learning with Curriculum Sampling (RLCS) to significantly enhance its capabilities in complex tasks. As a 9B-parameter model, it achieves state-of-the-art performance among models of a similar size, and its performance is comparable to or even surpasses the much larger 72B-parameter Qwen-2.5-VL-72B on 18 different benchmarks. The model excels in a diverse range of tasks, including STEM problem-solving, video understanding, and long document understanding, and it can handle images with resolutions up to 4K and arbitrary aspect ratios...

GLM-4.5 is a foundational model specifically designed for AI agent applications, built on a Mixture-of-Experts (MoE) architecture. It has been extensively optimized for tool use, web browsing, software development, and front-end development, enabling seamless integration with coding agents such as Claude Code and Roo Code. GLM-4.5 employs a hybrid reasoning approach, allowing it to adapt effectively to a wide range of application scenarios—from complex reasoning tasks to everyday use cases...

GLM-4.5-Air is a foundational model specifically designed for AI agent applications, built on a Mixture-of-Experts (MoE) architecture. It has been extensively optimized for tool use, web browsing, software development, and front-end development, enabling seamless integration with coding agents such as Claude Code and Roo Code. GLM-4.5 employs a hybrid reasoning approach, allowing it to adapt effectively to a wide range of application scenarios—from complex reasoning tasks to everyday use cases...

GLM-4.5V is the latest generation vision-language model (VLM) released by Zhipu AI. The model is built upon the flagship text model GLM-4.5-Air, which has 106B total parameters and 12B active parameters, and it utilizes a Mixture-of-Experts (MoE) architecture to achieve superior performance at a lower inference cost. Technically, GLM-4.5V follows the lineage of GLM-4.1V-Thinking and introduces innovations like 3D Rotated Positional Encoding (3D-RoPE), significantly enhancing its perception and reasoning abilities for 3D spatial relationships. Through optimization across pre-training, supervised fine-tuning, and reinforcement learning phases, the model is capable of processing diverse visual content such as images, videos, and long documents, achieving state-of-the-art performance among open-source models of its scale on 41 public multimodal benchmarks. Additionally, the model features a 'Thinking Mode' switch, allowing users to flexibly choose between quick responses and deep reasoning to balance efficiency and effectiveness...

GLM-Z1-32B-0414 is a reasoning model with deep thinking capabilities. This model was developed based on GLM-4-32B-0414 through cold start and extended reinforcement learning, as well as further training on tasks involving mathematics, code, and logic. Compared to the base model, GLM-Z1-32B-0414 significantly improves mathematical abilities and the capability to solve complex tasks. During the training process, the team also introduced general reinforcement learning based on pairwise ranking feedback, further enhancing the model's general capabilities. Despite having only 32B parameters, its performance on certain tasks is comparable to DeepSeek-R1 with 671B parameters. Through evaluations on benchmarks such as AIME 24/25, LiveCodeBench, and GPQA, the model demonstrates strong mathematical reasoning abilities and can support solutions for a wider range of complex tasks...

GLM-Z1-9B-0414 is a small-sized model in the GLM series with only 9 billion parameters that maintains the open-source tradition while showcasing surprising capabilities. Despite its smaller scale, GLM-Z1-9B-0414 still exhibits excellent performance in mathematical reasoning and general tasks. Its overall performance is already at a leading level among open-source models of the same size. The research team employed the same series of techniques used for larger models to train this 9B model. Especially in resource-constrained scenarios, this model achieves an excellent balance between efficiency and effectiveness, providing a powerful option for users seeking lightweight deployment. The model features deep thinking capabilities and can handle long contexts through YaRN technology, making it particularly suitable for applications requiring mathematical reasoning abilities with limited computational resources...

GLM-Z1-Rumination-32B-0414 is a deep reasoning model with rumination capabilities (benchmarked against OpenAI's Deep Research). Unlike typical deep thinking models, the rumination model employs longer periods of deep thought to solve more open-ended and complex problems (e.g., writing a comparative analysis of AI development in two cities and their future development plans). The rumination model integrates search tools during its deep thinking process to handle complex tasks and is trained by utilizing multiple rule-based rewards to guide and extend end-to-end reinforcement learning. Z1-Rumination shows significant improvements in research-style writing and complex retrieval tasks. The model supports a complete research cycle of “independently raising questions—searching for information—building analysis—completing tasks” and includes function calls like search, click, open, and finish by default, enabling it to better handle complex problems that require external information...

Hunyuan-A13B-Instruct activates only 13 B of its 80 B parameters, yet matches much larger LLMs on mainstream benchmarks. It offers hybrid reasoning: low-latency “fast” mode or high-precision “slow” mode, switchable per call. Native 256 K-token context lets it digest book-length documents without degradation. Agent skills are tuned for BFCL-v3, τ-Bench and C3-Bench leadership, making it an excellent autonomous assistant backbone. Grouped Query Attention plus multi-format quantization delivers memory-light, GPU-efficient inference for real-world deployment, with built-in multilingual support and robust safety alignment for enterprise-grade applications....

The Hunyuan Translation Model consists of a translation model, Hunyuan-MT-7B, and an ensemble model, Hunyuan-MT-Chimera. Hunyuan-MT-7B is a lightweight translation model with 7 billion parameters used to translate source text into the target language. The model supports mutual translation among 33 languages, including five ethnic minority languages in China. In the WMT25 machine translation competition, Hunyuan-MT-7B won first place in 30 out of the 31 language categories it participated in, demonstrating its outstanding translation capabilities. For translation tasks, Tencent Hunyuan proposed a comprehensive training framework covering pre-training, supervised fine-tuning, translation enhancement, and ensemble refinement, achieving state-of-the-art performance among models of a similar scale. The model is computationally efficient and easy to deploy, making it suitable for various application scenarios...

Kimi-Dev-72B is a new open-source coding large language model achieving 60.4% on SWE-bench Verified, setting a state-of-the-art result among open-source models. Optimized through large-scale reinforcement learning, it autonomously patches real codebases in Docker and earns rewards only when full test suites pass. This ensures the model delivers correct, robust, and practical solutions aligned with real-world software engineering standards...

Kimi K2 is a Mixture-of-Experts (MoE) foundation model with exceptional coding and agent capabilities, featuring 1 trillion total parameters and 32 billion activated parameters. In benchmark evaluations covering general knowledge reasoning, programming, mathematics, and agent-related tasks, the K2 model outperforms other leading open-source models...

Kimi K2-Instruct-0905 is the latest, most capable version of Kimi K2. It is a state-of-the-art mixture-of-experts (MoE) language model, featuring 32 billion activated parameters and a total of 1 trillion parameters. Key features include enhanced agentic coding intelligence, with the model demonstrating significant improvements on public benchmarks and real-world coding agent tasks; an improved frontend coding experience, offering advancements in both the aesthetics and practicality of frontend programming...

Ling-flash-2.0 is a language model from inclusionAI with a total of 100 billion parameters, of which 6.1 billion are activated per token (4.8 billion non-embedding). As part of the Ling 2.0 architecture series, it is designed as a lightweight yet powerful Mixture-of-Experts (MoE) model. It aims to deliver performance comparable to or even exceeding that of 40B-level dense models and other larger MoE models, but with a significantly smaller active parameter count. The model represents a strategy focused on achieving high performance and efficiency through extreme architectural design and training methods...

Ling-mini-2.0 is a small yet high-performance large language model built on the MoE architecture. It has 16B total parameters, but only 1.4B are activated per token (non-embedding 789M), enabling extremely fast generation. Thanks to the efficient MoE design and large-scale high-quality training data, despite having only 1.4B activated parameters, Ling-mini-2.0 still delivers top-tier downstream task performance comparable to sub-10B dense LLMs and even larger MoE models...

Meta Llama 3.1 is a family of multilingual large language models developed by Meta, featuring pretrained and instruction-tuned variants in 8B, 70B, and 405B parameter sizes. This 8B instruction-tuned model is optimized for multilingual dialogue use cases and outperforms many available open-source and closed chat models on common industry benchmarks. The model was trained on over 15 trillion tokens of publicly available data, using techniques like supervised fine-tuning and reinforcement learning with human feedback to enhance helpfulness and safety. Llama 3.1 supports text and code generation, with a knowledge cutoff of December 2023...

MiniMax-M1 is a open-weight, large-scale hybrid-attention reasoning model with 456 B parameters and 45.9 B activated per token. It natively supports 1 M-token context, lightning attention enabling 75% FLOPs savings vs DeepSeek R1 at 100 K tokens, and leverages a MoE architecture. Efficient RL training with CISPO and hybrid design yields state-of-the-art performance on long-input reasoning and real-world software engineering tasks....

QwQ is the reasoning model of the Qwen series. Compared with conventional instruction-tuned models, QwQ, which is capable of thinking and reasoning, can achieve significantly enhanced performance in downstream tasks, especially hard problems. QwQ-32B is the medium-sized reasoning model, which is capable of achieving competitive performance against state-of-the-art reasoning models, e.g., DeepSeek-R1, o1-mini. The model incorporates technologies like RoPE, SwiGLU, RMSNorm, and Attention QKV bias, with 64 layers and 40 Q attention heads (8 for KV in GQA architecture)...

Qwen2.5-14B-Instruct is one of the latest large language model series released by Alibaba Cloud. This 14B model demonstrates significant improvements in areas such as coding and mathematics. The model also offers multi-language support, covering over 29 languages, including Chinese and English. It has shown notable advancements in instruction following, understanding structured data, and generating structured outputs, particularly in JSON format....

Qwen2.5-32B-Instruct is one of the latest large language models series released by Alibaba Cloud. This 32B model demonstrates significant improvements in areas such as coding and mathematics. The model also offers multi-language support, covering over 29 languages, including Chinese, English, and others. It shows notable enhancements in instruction following, understanding structured data, and generating structured outputs, particularly in JSON format....

Qwen2.5-72B-Instruct is one of the latest large language models series released by Alibaba Cloud. This 72B model demonstrates significant improvements in areas such as coding and mathematics. It supports a context length of up to 128K tokens. The model also offers multilingual support, covering over 29 languages, including Chinese, English, and others. It has shown notable enhancements in instruction following, understanding structured data, and generating structured outputs, particularly in JSON format....

Qwen2.5-7B-Instruct is one of the latest large language model series released by Alibaba Cloud. This 7B model demonstrates significant improvements in areas such as coding and mathematics. The model also offers multilingual support, covering over 29 languages, including Chinese, English, and others. The model shows notable enhancements in instruction following, understanding structured data, and generating structured outputs, particularly JSON....

Qwen2.5-Coder-32B-Instruct is a code-specific large language model developed based on Qwen2.5. The model has undergone training on 5.5 trillion tokens, achieving significant improvements in code generation, code reasoning, and code repair. It is currently the most advanced open-source code language model, with coding capabilities comparable to GPT-4. Not only has the model enhanced coding abilities, but it also maintains strengths in mathematics and general capabilities, and supports long text processing....

Qwen2.5-VL-32B-Instruct is a multimodal large language model released by the Qwen team, part of the Qwen2.5-VL series. This model is not only proficient in recognizing common objects but is highly capable of analyzing texts, charts, icons, graphics, and layouts within images. It acts as a visual agent that can reason and dynamically direct tools, capable of computer and phone use. Additionally, the model can accurately localize objects in images, and generate structured outputs for data like invoices and tables. Compared to its predecessor Qwen2-VL, this version has enhanced mathematical and problem-solving abilities through reinforcement learning, with response styles adjusted to better align with human preferences...

Qwen2.5-VL is a vision-language model in the Qwen2.5 series that shows significant enhancements in several aspects: it has strong visual understanding capabilities, recognizing common objects while analyzing texts, charts, and layouts in images; it functions as a visual agent capable of reasoning and dynamically directing tools; it can comprehend videos over 1 hour long and capture key events; it accurately localizes objects in images by generating bounding boxes or points; and it supports structured outputs for scanned data like invoices and forms. The model demonstrates excellent performance across various benchmarks including image, video, and agent tasks...

Qwen2.5-VL is a new member of the Qwen series, equipped with powerful visual comprehension capabilities. It can analyze text, charts, and layouts within images, understand long videos, and capture events. It is capable of reasoning, manipulating tools, supporting multi-format object localization, and generating structured outputs. The model has been optimized for dynamic resolution and frame rate training in video understanding, and has improved the efficiency of the visual encoder....

Qwen3-14B is the latest large language model in the Qwen series with 14.8B parameters. This model uniquely supports seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue). It demonstrates significantly enhanced reasoning capabilities, surpassing previous QwQ and Qwen2.5 instruct models in mathematics, code generation, and commonsense logical reasoning. The model excels in human preference alignment for creative writing, role-playing, and multi-turn dialogues. Additionally, it supports over 100 languages and dialects with strong multilingual instruction following and translation capabilities...

Qwen3-235B-A22B is the latest large language model in the Qwen series, featuring a Mixture-of-Experts (MoE) architecture with 235B total parameters and 22B activated parameters. This model uniquely supports seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue). It demonstrates significantly enhanced reasoning capabilities, superior human preference alignment in creative writing, role-playing, and multi-turn dialogues. The model excels in agent capabilities for precise integration with external tools and supports over 100 languages and dialects with strong multilingual instruction following and translation capabilities...

Qwen3-235B-A22B-Instruct-2507 is a flagship Mixture-of-Experts (MoE) large language model from the Qwen3 series, developed by Alibaba Cloud's Qwen team. The model has a total of 235 billion parameters, with 22 billion activated per forward pass. It was released as an updated version of the Qwen3-235B-A22B non-thinking mode, featuring significant enhancements in general capabilities such as instruction following, logical reasoning, text comprehension, mathematics, science, coding, and tool usage. Additionally, the model provides substantial gains in long-tail knowledge coverage across multiple languages and shows markedly better alignment with user preferences in subjective and open-ended tasks, enabling more helpful responses and higher-quality text generation. Notably, it natively supports an extensive 256K (262,144 tokens) context window, which enhances its capabilities for long-context understanding. This version exclusively supports the non-thinking mode and does not generate <think> blocks, aiming to deliver more efficient and precise responses for tasks like direct Q&A and knowledge retrieval...

Qwen3-235B-A22B-Thinking-2507 is a member of the Qwen3 large language model series developed by Alibaba's Qwen team, specializing in highly complex reasoning tasks. The model is built on a Mixture-of-Experts (MoE) architecture, with 235 billion total parameters and approximately 22 billion activated parameters per token, which enhances computational efficiency while maintaining powerful performance. As a dedicated 'thinking' model, it demonstrates significantly improved performance on tasks requiring human expertise, such as logical reasoning, mathematics, science, coding, and academic benchmarks, achieving state-of-the-art results among open-source thinking models. Furthermore, the model features enhanced general capabilities like instruction following, tool usage, and text generation, and it natively supports a 256K long-context understanding capability, making it ideal for scenarios that require deep reasoning and processing of long documents...

Qwen3-30B-A3B is the latest large language model in the Qwen series, featuring a Mixture-of-Experts (MoE) architecture with 30.5B total parameters and 3.3B activated parameters. This model uniquely supports seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue). It demonstrates significantly enhanced reasoning capabilities, superior human preference alignment in creative writing, role-playing, and multi-turn dialogues. The model excels in agent capabilities for precise integration with external tools and supports over 100 languages and dialects with strong multilingual instruction following and translation capabilities...

Qwen3-30B-A3B-Instruct-2507 is the updated version of the Qwen3-30B-A3B non-thinking mode. It is a Mixture-of-Experts (MoE) model with 30.5 billion total parameters and 3.3 billion activated parameters. This version features key enhancements, including significant improvements in general capabilities such as instruction following, logical reasoning, text comprehension, mathematics, science, coding, and tool usage. It also shows substantial gains in long-tail knowledge coverage across multiple languages and offers markedly better alignment with user preferences in subjective and open-ended tasks, enabling more helpful responses and higher-quality text generation. Furthermore, its capabilities in long-context understanding have been enhanced to 256K. This model supports only non-thinking mode and does not generate `<think></think>` blocks in its output...

Qwen3-30B-A3B-Thinking-2507 is the latest thinking model in the Qwen3 series, released by Alibaba's Qwen team. As a Mixture-of-Experts (MoE) model with 30.5 billion total parameters and 3.3 billion active parameters, it is focused on enhancing capabilities for complex tasks. The model demonstrates significantly improved performance on reasoning tasks, including logical reasoning, mathematics, science, coding, and academic benchmarks that typically require human expertise. It also shows markedly better general capabilities, such as instruction following, tool usage, text generation, and alignment with human preferences. The model natively supports a 256K long-context understanding capability, which can be extended to 1 million tokens. This version is specifically designed for ‘thinking mode’ to tackle highly complex problems through step-by-step reasoning and also excels in agentic capabilities...

Qwen3-32B is the latest large language model in the Qwen series with 32.8B parameters. This model uniquely supports seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue). It demonstrates significantly enhanced reasoning capabilities, surpassing previous QwQ and Qwen2.5 instruct models in mathematics, code generation, and commonsense logical reasoning. The model excels in human preference alignment for creative writing, role-playing, and multi-turn dialogues. Additionally, it supports over 100 languages and dialects with strong multilingual instruction following and translation capabilities...

Qwen3-8B is the latest large language model in the Qwen series with 8.2B parameters. This model uniquely supports seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue). It demonstrates significantly enhanced reasoning capabilities, surpassing previous QwQ and Qwen2.5 instruct models in mathematics, code generation, and commonsense logical reasoning. The model excels in human preference alignment for creative writing, role-playing, and multi-turn dialogues. Additionally, it supports over 100 languages and dialects with strong multilingual instruction following and translation capabilities...

Qwen3-Coder-30B-A3B-Instruct is a code model from the Qwen3 series developed by Alibaba's Qwen team. As a streamlined and optimized model, it maintains impressive performance and efficiency while focusing on enhanced coding capabilities. It demonstrates significant performance advantages among open-source models on complex tasks such as Agentic Coding, Agentic Browser-Use, and other foundational coding tasks. The model natively supports a long context of 256K tokens, which can be extended up to 1M tokens, enabling better repository-scale understanding and processing. Furthermore, it provides robust agentic coding support for platforms like Qwen Code and CLINE, featuring a specially designed function call format...

Qwen3-Coder-480B-A35B-Instruct is the most agentic code model released by Alibaba to date. It is a Mixture-of-Experts (MoE) model with 480 billion total parameters and 35 billion activated parameters, balancing efficiency and performance. The model natively supports a 256K (approximately 262,144) token context length, which can be extended up to 1 million tokens using extrapolation methods like YaRN, enabling it to handle repository-scale codebases and complex programming tasks. Qwen3-Coder is specifically designed for agentic coding workflows, where it not only generates code but also autonomously interacts with developer tools and environments to solve complex problems. It has achieved state-of-the-art results among open models on various coding and agentic benchmarks, with performance comparable to leading models like Claude Sonnet 4. Alongside the model, Alibaba has also open-sourced Qwen Code, a command-line tool designed to fully unleash its powerful agentic coding capabilities...

Qwen3-Embedding-0.6B is the latest proprietary model in the Qwen3 Embedding series, specifically designed for text embedding and ranking tasks. Built upon the dense foundational models of the Qwen3 series, this 0.6B parameter model supports context lengths up to 32K and can generate embeddings with dimensions up to 1024. The model inherits exceptional multilingual capabilities supporting over 100 languages, along with long-text understanding and reasoning skills. It achieves strong performance on the MTEB multilingual leaderboard (score 64.33) and demonstrates excellent results across various tasks including text retrieval, code retrieval, text classification, clustering, and bitext mining. The model offers flexible vector dimensions (32 to 1024) and instruction-aware capabilities for enhanced performance in specific tasks and scenarios, making it an ideal choice for applications prioritizing both efficiency and effectiveness...

Qwen3-Embedding-4B is the latest proprietary model in the Qwen3 Embedding series, specifically designed for text embedding and ranking tasks. Built upon the dense foundational models of the Qwen3 series, this 4B parameter model supports context lengths up to 32K and can generate embeddings with dimensions up to 2560. The model inherits exceptional multilingual capabilities supporting over 100 languages, along with long-text understanding and reasoning skills. It achieves excellent performance on the MTEB multilingual leaderboard (score 69.45) and demonstrates outstanding results across various tasks including text retrieval, code retrieval, text classification, clustering, and bitext mining. The model offers flexible vector dimensions (32 to 2560) and instruction-aware capabilities for enhanced performance in specific tasks and scenarios, providing an optimal balance between efficiency and effectiveness...

Qwen3-Embedding-8B is the latest proprietary model in the Qwen3 Embedding series, specifically designed for text embedding and ranking tasks. Built upon the dense foundational models of the Qwen3 series, this 8B parameter model supports context lengths up to 32K and can generate embeddings with dimensions up to 4096. The model inherits exceptional multilingual capabilities supporting over 100 languages, along with long-text understanding and reasoning skills. It ranks No.1 on the MTEB multilingual leaderboard (as of June 5, 2025, score 70.58) and demonstrates state-of-the-art performance across various tasks including text retrieval, code retrieval, text classification, clustering, and bitext mining. The model offers flexible vector dimensions (32 to 4096) and instruction-aware capabilities for enhanced performance in specific tasks and scenarios...

Qwen3-Next-80B-A3B-Instruct is a next-generation foundation model released by Alibaba's Qwen team. It is built on the new Qwen3-Next architecture, designed for ultimate training and inference efficiency. The model incorporates innovative features such as a Hybrid Attention mechanism (Gated DeltaNet and Gated Attention), a High-Sparsity Mixture-of-Experts (MoE) structure, and various stability optimizations. As an 80-billion-parameter sparse model, it activates only about 3 billion parameters per token during inference, which significantly reduces computational costs and delivers over 10 times higher throughput than the Qwen3-32B model for long-context tasks exceeding 32K tokens. This is an instruction-tuned version optimized for general-purpose tasks and does not support 'thinking' mode. In terms of performance, it is comparable to Qwen's flagship model, Qwen3-235B, on certain benchmarks, showing significant advantages in ultra-long-context scenarios...

Qwen3-Reranker-0.6B is a text reranking model from the Qwen3 series. It is specifically designed to refine the results from initial retrieval systems by re-ordering documents based on their relevance to a given query. With 0.6 billion parameters and a context length of 32k, this model leverages the strong multilingual (supporting over 100 languages), long-text understanding, and reasoning capabilities of its Qwen3 foundation. Evaluation results show that Qwen3-Reranker-0.6B achieves strong performance across various text retrieval benchmarks, including MTEB-R, CMTEB-R, and MLDR...

Qwen3-Reranker-4B is a powerful text reranking model from the Qwen3 series, featuring 4 billion parameters. It is engineered to significantly improve the relevance of search results by re-ordering an initial list of documents based on a query. This model inherits the core strengths of its Qwen3 foundation, including exceptional understanding of long-text (up to 32k context length) and robust capabilities across more than 100 languages. According to benchmarks, the Qwen3-Reranker-4B model demonstrates superior performance in various text and code retrieval evaluations...

Qwen3-Reranker-8B is the 8-billion parameter text reranking model from the Qwen3 series. It is designed to refine and improve the quality of search results by accurately re-ordering documents based on their relevance to a query. Built on the powerful Qwen3 foundational models, it excels in understanding long-text with a 32k context length and supports over 100 languages. The Qwen3-Reranker-8B model is part of a flexible series that offers state-of-the-art performance in various text and code retrieval scenarios...

Seed-OSS is a series of open-source large language models developed by the ByteDance Seed team, designed for powerful long-context processing, reasoning, agent capabilities, and general-purpose abilities. Within this series, Seed-OSS-36B-Instruct is an instruction-tuned model with 36 billion parameters that natively supports an ultra-long context length, enabling it to process massive documents or complex codebases in a single pass. The model is specially optimized for reasoning, code generation, and agent tasks (such as tool use), while maintaining balanced and excellent general-purpose capabilities. A key feature of this model is the ‘Thinking Budget’ function, which allows users to flexibly adjust the reasoning length as needed, thereby effectively improving inference efficiency in practical applications...

gpt-oss-120b is OpenAI’s open-weight large language model with ~117B parameters (5.1B active), using a Mixture-of-Experts (MoE) design and MXFP4 quantization to run on a single 80 GB GPU. It delivers o4-mini-level or better performance in reasoning, coding, health, and math benchmarks, with full Chain-of-Thought (CoT), tool use, and Apache 2.0-licensed commercial deployment support....

gpt-oss-20b is OpenAI’s lightweight open-weight model with ~21B parameters (3.6B active), built on an MoE architecture and MXFP4 quantization to run locally on 16 GB VRAM devices. It matches o3-mini in reasoning, math, and health tasks, supporting CoT, tool use, and deployment via frameworks like Transformers, vLLM, and Ollama....

Step3 is a cutting-edge multimodal reasoning model from StepFun. It is built on a Mixture-of-Experts (MoE) architecture with 321B total parameters and 38B active parameters. The model is designed end-to-end to minimize decoding costs while delivering top-tier performance in vision-language reasoning. Through the co-design of Multi-Matrix Factorization Attention (MFA) and Attention-FFN Disaggregation (AFD), Step3 maintains exceptional efficiency across both flagship and low-end accelerators. During pretraining, Step3 processed over 20T text tokens and 4T image-text mixed tokens, spanning more than ten languages. The model has achieved state-of-the-art performance for open-source models on various benchmarks, including math, code, and multimodality...

Llama 3.3 is the most advanced multilingual open-source large language model in the Llama series, offering performance comparable to a 405B model at a significantly lower cost. Built on the Transformer architecture, it enhances usefulness and safety through supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF). Its instruction-tuned version is optimized for multilingual dialogue and outperforms many open-source and closed chat models across various industry benchmarks. The knowledge cutoff is December 2023....

Qwen2.5-72B-Instruct is one of the latest large language model series released by Alibaba Cloud. The 72B model demonstrates significant improvements in areas such as coding and mathematics. The model also offers multilingual support, covering over 29 languages, including Chinese and English. It shows notable enhancements in following instructions, understanding structured data, and generating structured outputs, particularly in JSON format....