浏览 AI 模型
共 374 个模型s可用
MistralDevstral 2 123B InstructDevstral is an agentic LLM for software engineering tasks. Devstral 2 excels at using tools to explore codebases, editing multiple files and power software engineering agents. The model achieves remarkable performance on SWE-bench.
Moonshot AIKimi K2.5Kimi K2.5 is Moonshot AI's advanced reasoning model with strong performance in math, coding, and multilingual tasks. Features long-context understanding and agentic capabilities for complex multi-step problem solving.
Moonshot AIKimi K2.6Kimi K2.6 is Moonshot AI's open-weight multimodal agentic model, focused on long-horizon coding, coding-driven design, autonomous execution, and swarm-style task orchestration.
AlibabaQwen 3.5 397B A17BQwen3.5 397B A17B is the flagship of the Qwen3.5 family — a massive MoE model with frontier-level quality across all tasks.
AlibabaQwen3-Coder 30B A3B InstructQwen3-Coder is available in multiple sizes. Today, we're excited to introduce Qwen3-Coder-30B-A3B-Instruct. This streamlined model maintains impressive performance and efficiency, featuring the following key enhancements:
DeepSeekDeepSeek V4 ProDeepSeek V4 Pro is a 1.6T-parameter sparse MoE (49B active, 384 routed + 1 shared expert) built for million-token agentic reasoning. Experts ship natively in FP4, so the real on-disk footprint is roughly 862 GB (FP4 experts + FP8 attention) rather than the trillion-scale FP16 size — but it is still a server/workstation deployment: realistic local use targets 8x 80GB GPUs or 1 TB+ unified memory, and at long Think Max contexts the KV cache dominates.
AlibabaQwen 3.5 27BQwen3.5 27B provides frontier-level quality in a size that fits on high-end consumer GPUs. Excellent for coding, reasoning, and RAG workloads.
AlibabaQwen 3.6 27BQwen 3.6 27B is a dense, multimodal 27B model that beats the previous-gen Qwen 3.5 397B-A17B MoE flagship on SWE-bench Verified (77.2% vs 76.2%) while fitting on a 16-24 GB consumer GPU. It uses a Gated DeltaNet + Gated Attention hybrid, ships a vision encoder, and supports 262K native context extensible to ~1M tokens via YaRN.
AlibabaQwen 3.5 122B A10BQwen3.5 122B A10B is a high-quality MoE model with 10B active parameters. Strong performance across coding, reasoning, and RAG with moderate inference cost.
DeepSeekDeepSeek V4 FlashDeepSeek V4 Flash is the lighter 284B-parameter sparse MoE sibling of V4 Pro (13B active, 256 routed + 1 shared expert) with the same 1M-token context. Experts ship natively in FP4, so the real on-disk footprint is roughly 158 GB rather than the FP16 size — it fits a single 192 GB unified-memory machine or a 2-4 GPU server while keeping near-frontier reasoning and coding quality.
AlibabaQwen 3.6 35B A3BQwen 3.6 35B A3B is the first open-weight Qwen 3.6 model, a multimodal MoE release focused on stronger agentic coding, long-context reasoning, and more stable repository-scale workflows.
AlibabaQwen3-VL 30B A3B InstructMeet Qwen3-VL — the most powerful vision-language model in the Qwen series to date.
AlibabaQwen 3.5 9BQwen3.5 9B is a strong all-rounder at the popular 7-9B parameter sweet spot, with significant improvements in reasoning, code generation, and instruction following.
AlibabaQwen 3.5 35B A3BQwen3.5 35B A3B is a Mixture-of-Experts model with only 3B active parameters per token, offering surprisingly strong performance at very low inference cost.
MistralMagistral Small 2507Building upon Mistral Small 3.1 (2503), with added reasoning capabilities, undergoing SFT from Magistral Medium traces and RL on top, it's a small, efficient reasoning model with 24B parameters.
MistralDevstral Small 2 24B InstructDevstral is an agentic LLM for software engineering tasks. Devstral Small 2 excels at using tools to explore codebases, editing multiple files and power software engineering agents. The model achieves remarkable performance on SWE-bench.
AlibabaQwen 3 32BQwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features:
AlibabaQwen 3 14BQwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features:
MistralMistral Small 4 119BMistral Small 4 is a powerful hybrid model capable of acting as both a general instruction model and a reasoning model. It unifies the capabilities of three different model families—Instruct, Reasoning (previously called Magistral), and Devstral—into a single, unified model.
AlibabaQwen 3 30B A3BWe introduce the updated version of the Qwen3-30B-A3B non-thinking mode, named Qwen3-30B-A3B-Instruct-2507, featuring the following key enhancements:
CohereCommand A 111BCommand A is Cohere's latest flagship model with 111B parameters, designed for agentic enterprise applications. Features advanced tool use, multi-step reasoning, and retrieval-augmented generation.
OpenAIGPT-OSS 120BGPT-OSS 120B is OpenAI's large open-source model, offering strong reasoning and coding capabilities.
AlibabaQwen 2.5 VL 72Blicense: other license_name: qwen license_link: https://huggingface.co/Qwen/Qwen2.5-VL-72B-Instruct/blob/main/LICENSE language: - en pipeline_tag: image-text-to-text tags: - multimodal library_name: transformers
NVIDIANemotron 3 Nano 30BNemotron 3 Nano 30B is NVIDIA's mid-size reasoning model delivering strong performance across coding, math, and agentic tasks. Fits on a 24 GB GPU at Q4_K_M.