Xinference 内置 LLM 模型目录163 个模型家族的规格、能力与启动参数选择【免费下载链接】inferenceSwap GPT for any LLM by changing a single line of code. Xinference lets you run open-source, speech, and multimodal models on cloud, on-prem, or your laptop — all through one unified, production-ready inference API.项目地址: https://gitcode.com/GitHub_Trending/in/inference本文以官方文档 doc/source/models/builtin/llm/index.rst 为主体完整收录 Xinference 内置 LLM 目录页的模型清单模型名、能力、上下文长度、描述并结合文档生成器 gen_docs.py、模型注册表 llm_family.json 与 Jinja 模板讲清目录页每个字段的来源以及如何把目录中的一行模型记录翻译成可执行的xinference launch启动命令。读完后你可以按能力与上下文长度筛选模型、读懂每个模型详情页的规格参数、为指定量化/格式选择正确的启动参数。目录页在文档体系中的位置与内容结构Xinference 的模型文档入口是 doc/source/models/index.rst它给出“列出模型 / 启动与终止模型 / 模型使用”三类操作并通过 toctree 进入 Builtin Models 下的七个类别llm、embedding、image、audio、rerank、video、world。本文聚焦其中的 LLM 类别。LLM 索引页 index.rst 由两部分构成四列汇总表格MODEL NAME / ABILITIES / COTNEXT_LENGTH / DESCRIPTION原文表头如此拼写即 Context Length一行对应一个内置模型家族模型名是一个指向详情页的引用:ref:角色链接toctree 目录树把每个模型名展开为独立子页面doc/source/models/builtin/llm/模型名.rst共 163 个模型家族例如 deepseek-r1、qwen3、glm-5。每个模型详情页的字段结构由模板 doc/templates/llm.rst.jinja 统一生成包含字段含义Context Length模型声明的上下文长度token 数Model Name用于--model-name参数的模型名Languages模型支持的语言Abilities模型能力标签见下文词汇表Description模型介绍来自注册表原文每组 Model SpecModel Formatpytorch / awq / ggufv2 / mlx 等、Model Size以十亿参数计、Quantizations可选量化列表、Engines支持的后端引擎、Model ID模型仓库 ID、Model HubsHugging Face / ModelScope以及对应的xinference launch命令能力词汇表ABILITIES 列取值按目录页实际出现的组合归纳generate基础文本生成未做对话对齐的 base 模型chat对话模型支持 chat 接口reasoning推理/思维链模型如 deepseek-r1、qwq-32bhybrid同时支持思考与非思考模式——从目录描述可以印证如deepseek-v3.1的 Description 写明 “supports both thinking mode and non-thinking mode”minicpm5-1b写明 “Supports hybrid thinking via enable_thinking”tools支持工具调用vision/audio/omni视觉、音频、全模态输入各能力的完整语义说明可参考 doc/source/models/model_abilities/ 目录下的专题文档。上下文长度分布可直接用于选型目录中最小为gpt-2的 1024最大为 10485761M token对应kimi-k3、glm-5.2、glm4-chat-1m、minimax-m3、deepseek-v4-flash-0731五个模型mistral-nemo-instruct1024000与qwen2.5-instruct-1m1010000也接近 1M。大量旗舰模型处于 131072262144 区间128K256K。内置 LLM 完整清单以下为索引页表格的完整内容163 行按原文顺序。模型名均对应 doc/source/models/builtin/llm/ 目录下的同名.rst详情页。Model NameAbilitiesContext LengthDescriptionbaichuan-2generate4096Baichuan2 is an open-source Transformer based LLM that is trained on both Chinese and English data.baichuan-2-chatchat4096Baichuan2-chat is a fine-tuned version of the Baichuan LLM, specializing in chatting.baichuan-m2chat, reasoning, hybrid, tools131072Baichuan-M2-32B is Baichuan AIs medical-enhanced reasoning model, the second medical model released by Baichuan. Designed for real-world medical reasoning tasks, this model builds upon Qwen2.5-32B with an innovative Large Verifier System. Through domain-specific fine-tuning on real-world medical questions, it achieves breakthrough medical performance while maintaining strong general capabilities.code-llamagenerate100000Code-Llama is an open-source LLM trained by fine-tuning LLaMA2 for generating and discussing code.code-llama-instructchat100000Code-Llama-Instruct is an instruct-tuned version of the Code-Llama LLM.code-llama-pythongenerate100000Code-Llama-Python is a fine-tuned version of the Code-Llama LLM, specializing in Python.codegeex4chat131072the open-source version of the latest CodeGeeX4 model seriescodeqwen1.5generate65536CodeQwen1.5 is the Code-Specific version of Qwen1.5. It is a transformer-based decoder-only language model pretrained on a large amount of data of codes.codeqwen1.5-chatchat65536CodeQwen1.5 is the Code-Specific version of Qwen1.5. It is a transformer-based decoder-only language model pretrained on a large amount of data of codes.codestral-v0.1generate32768Codestrall-22B-v0.1 is trained on a diverse dataset of 80 programming languages, including the most popular ones, such as Python, Java, C, C, JavaScript, and Bashcogagentchat, vision4096The CogAgent-9B-20241220 model is based on GLM-4V-9B, a bilingual open-source VLM base model. Through data collection and optimization, multi-stage training, and strategy improvements, CogAgent-9B-20241220 achieves significant advancements in GUI perception, inference prediction accuracy, action space completeness, and task generalizability.deepseekgenerate4096DeepSeek LLM, trained from scratch on a vast dataset of 2 trillion tokens in both English and Chinese.deepseek-chatchat4096DeepSeek LLM is an advanced language model comprising 67 billion parameters. It has been trained from scratch on a vast dataset of 2 trillion tokens in both English and Chinese.deepseek-codergenerate16384Deepseek Coder is composed of a series of code language models, each trained from scratch on 2T tokens, with a composition of 87% code and 13% natural language in both English and Chinese.deepseek-coder-instructchat16384deepseek-coder-instruct is a model initialized from deepseek-coder-base and fine-tuned on 2B tokens of instruction data.deepseek-prover-v2chat, reasoning163840We introduce DeepSeek-Prover-V2, an open-source large language model designed for formal theorem proving in Lean 4, with initialization data collected through a recursive theorem proving pipeline powered by DeepSeek-V3. The cold-start training procedure begins by prompting DeepSeek-V3 to decompose complex problems into a series of subgoals. The proofs of resolved subgoals are synthesized into a chain-of-thought process, combined with DeepSeek-V3s step-by-step reasoning, to create an initial cold start for reinforcement learning. This process enables us to integrate both informal and formal mathematical reasoning into a unified modeldeepseek-r1chat, reasoning163840DeepSeek-R1, which incorporates cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks.deepseek-r1-0528chat, reasoning, tools163840DeepSeek-R1, which incorporates cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks.deepseek-r1-0528-qwen3chat, reasoning131072The DeepSeek R1 model has undergone a minor version upgrade, with the current version being DeepSeek-R1-0528. In the latest update, DeepSeek R1 has significantly improved its depth of reasoning and inference capabilities by leveraging increased computational resources and introducing algorithmic optimization mechanisms during post-training. The model has demonstrated outstanding performance across various benchmark evaluations, including mathematics, programming, and general logic. Its overall performance is now approaching that of leading models, such as O3 and Gemini 2.5 Prodeepseek-r1-distill-llamachat, reasoning131072deepseek-r1-distill-llama is distilled from DeepSeek-R1 based on Llamadeepseek-r1-distill-qwenchat, reasoning131072deepseek-r1-distill-qwen is distilled from DeepSeek-R1 based on Qwendeepseek-v2-chatchat128000DeepSeek-V2, a strong Mixture-of-Experts (MoE) language model characterized by economical training and efficient inference.deepseek-v2-chat-0628chat128000DeepSeek-V2-Chat-0628 is an improved version of DeepSeek-V2-Chat.deepseek-v2.5chat128000DeepSeek-V2.5 is an upgraded version that combines DeepSeek-V2-Chat and DeepSeek-Coder-V2-Instruct. The new model integrates the general and coding abilities of the two previous versions.deepseek-v3chat163840DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token.deepseek-v3-0324chat163840DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token.deepseek-v3.1chat, reasoning, hybrid, tools131072DeepSeek-V3.1 is a hybrid model that supports both thinking mode and non-thinking mode.deepseek-v3.2chat, reasoning, hybrid, tools163840We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performancedeepseek-v3.2-expchat, reasoning, hybrid, tools163840We are excited to announce the official release of DeepSeek-V3.2-Exp, an experimental version of our model. As an intermediate step toward our next-generation architecture, V3.2-Exp builds upon V3.1-Terminus by introducing DeepSeek Sparse Attention—a sparse attention mechanism designed to explore and validate optimizations for training and inference efficiency in long-context scenarios.deepseek-v4-flashchat, reasoning, hybrid, tools163840We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models — DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) — both supporting a context length of one million tokens.deepseek-v4-flash-0731chat, reasoning, hybrid, tools1048576Official DeepSeek-V4-Flash release with enhanced agentic capabilities and an attached DSpark speculative decoding module.deepseek-v4-prochat, reasoning, hybrid, tools163840We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models — DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) — both supporting a context length of one million tokens.deepseek-vl2chat, vision4096DeepSeek-VL2, an advanced series of large Mixture-of-Experts (MoE) Vision-Language Models that significantly improves upon its predecessor, DeepSeek-VL. DeepSeek-VL2 demonstrates superior capabilities across various tasks, including but not limited to visual question answering, optical character recognition, document/table/chart understanding, and visual grounding.dianjin-r1chat, reasoning, hybrid, tools32768Tongyi DianJin is a financial intelligence solution platform built by Alibaba Cloud, dedicated to providing financial business developers with a convenient artificial intelligence application development environment.ernie4.5chat131072ERNIE 4.5, a new family of large-scale multimodal models comprising 10 distinct variants.fin-r1chat131072Fin-R1 is a large language model specifically designed for the field of financial reasoninggemma-3-1b-itchat32768Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models.gemma-3-itchat, vision131072Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models.gemma-4generate, chat, reasoning, tools, audio, vision, hybrid262144Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on small models) and generating text output.glm-4.1v-thinkingchat, vision, reasoning, tools65536GLM-4.1V-9B-Thinking, designed to explore the upper limits of reasoning in vision-language models.glm-4.5chat, reasoning, hybrid, tools131072The GLM-4.5 series models are foundation models designed for intelligent agents.glm-4.5vchat, vision, reasoning, tools131072GLM-4.5V is based on ZhipuAIs next-generation flagship text foundation model GLM-4.5-Air (106B parameters, 12B active). It continues the technical approach of GLM-4.1V-Thinking, achieving SOTA performance among models of the same scale on 42 public vision-language benchmarks.glm-4.6chat, reasoning, hybrid, tools202752GLM-4.6 significantly enhances context length (up to 200K tokens), code generation, reasoning with tool use, agent capabilities, and human-aligned writing compared to GLM-4.5.glm-4.7chat, reasoning, hybrid, tools202752GLM-4.7 significantly advances core and multilingual agentic coding, UI/vibe coding, tool use, and complex reasoning—outperforming GLM-4.6 across benchmarks like SWE-bench, Terminal Bench 2.0, τ²-Bench, and HLE—while also improving chat, creative writing, and role-play.glm-4.7-flashchat, reasoning, hybrid, tools202752GLM-4.7-Flash is a 30B-A3B MoE model. As the strongest model in the 30B class, it offers a lightweight deployment option that balances performance and efficiency.glm-4vchat, vision8192GLM4 is the open source version of the latest generation of pre-trained models in the GLM-4 series launched by Zhipu AI.glm-5chat, vision, tools, reasoning202752We are launching GLM-5, targeting complex systems engineering and long-horizon agentic tasks. Scaling is still one of the most important ways to improve the intelligence efficiency of Artificial General Intelligence (AGI). Compared to GLM-4.5, GLM-5 scales from 355B parameters (32B active) to 744B parameters (40B active), and increases pre-training data from 23T to 28.5T tokens. GLM-5 also integrates DeepSeek Sparse Attention (DSA), largely reducing deployment cost while preserving long-context capacity. Reinforcement learning aims to bridge the gap between competence and excellence in pre-trained models. However, deploying it at scale for LLMs is a challenge due to the RL training inefficiency. To this end, we developed slime, a novel asynchronous RL infrastructure that substantially improves training throughput and efficiency, enabling more fine-grained post-training iterations. With advances in both pre-training and post-training, GLM-5 delivers significant improvement compared to GLM-4.7 across a wide range of academic benchmarks and achieves best-in-class performance among all open-source models in the world on reasoning, coding, and agentic tasks, closing the gap with frontier models.glm-5.1chat, vision, tools, reasoning, hybrid202752GLM-5.1 is our next-generation flagship model for agentic engineering, with significantly stronger coding capabilities than its predecessor. It achieves state-of-the-art performance on SWE-Bench Pro and leads GLM-5 by a wide margin on NL2Repo (repo generation) and Terminal-Bench 2.0 (real-world terminal tasks).glm-5.2chat, tools, reasoning, hybrid1048576Were introducing GLM-5.2, our latest flagship model for long-horizon tasksglm-edge-chatchat8192The GLM-Edge series is our attempt to face the end-side real-life scenarios, which consists of two sizes of large-language dialogue models and multimodal comprehension models (GLM-Edge-1.5B-Chat, GLM-Edge-4B-Chat, GLM-Edge-V-2B, GLM-Edge-V-5B). Among them, the 1.5B / 2B model is mainly for platforms such as mobile phones and cars, and the 4B / 5B model is mainly for platforms such as PCs.glm4-0414chat, tools32768The GLM family welcomes new members, the GLM-4-32B-0414 series models, featuring 32 billion parameters. Its performance is comparable to OpenAIs GPT series and DeepSeeks V3/R1 seriesglm4-chatchat, tools131072GLM4 is the open source version of the latest generation of pre-trained models in the GLM-4 series launched by Zhipu AI.glm4-chat-1mchat, tools1048576GLM4 is the open source version of the latest generation of pre-trained models in the GLM-4 series launched by Zhipu AI.gpt-2generate1024GPT-2 is a Transformer-based LLM that is trained on WebTest, a 40 GB dataset of Reddit posts with 3 upvotes.gpt-osschat, reasoning131072gpt-oss series, OpenAIs open-weight models designed for powerful reasoning, agentic tasks, and versatile developer use cases【免费下载链接】inferenceSwap GPT for any LLM by changing a single line of code. Xinference lets you run open-source, speech, and multimodal models on cloud, on-prem, or your laptop — all through one unified, production-ready inference API.项目地址: https://gitcode.com/GitHub_Trending/in/inference创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考