Qwen3.5 · 35B A3B
Qwen3.5-35B-A3B is an Apache-2.0 licensed causal language model with a vision encoder, hosted in the Qwen/Qwen3.5-35B-A3B repository. It supports image-text-to-text generation and is provided in Hugging Face Transformers format with safetensors weights. The model has 35B total parameters with 3B activated, uses a hybrid architecture combining Gated DeltaNet, Gated Attention, and a sparse Mixture-of-Experts layer, and supports a native context length of 262,144 tokens, extensible up to 1,010,000 tokens. It is based on Qwen/Qwen3.5-35B-A3B-Base and is described as compatible with Hugging Face Transformers, vLLM, SGLang, and KTransformers. The model card highlights a unified vision-language foundation, efficient hybrid architecture, scalable reinforcement learning, coverage of 201 languages and dialects, and a next-generation training infrastructure.
At a glance
- Parameters
- 35B total · 3B active
- Architecture
- Mixture of experts
- Context length
- 262,144 native
- License
- Apache 2.0
- Disk space
- 24 GB
- Software
- Ollama, MLX, vLLM, SGLang
Will it run on your machine?
Save your machine to see a personalized rating and its reasoning.
Add my hardwareBest for
- Image-text-to-text generation
- Multimodal reasoning and visual understanding
- Coding tasks
- Agent-based applications
- Multilingual and cross-lingual generation across 201 languages and dialects
Tradeoffs
- No performance benchmarks, evaluations, or numerical hardware requirements are provided in the source facts.
- The source does not include information on licensing terms beyond Apache-2.0.
- Actual compatibility may depend on the specific inference framework and version; no detailed compatibility testing is described.
- No testing or validation of the model is claimed by the source facts.
- This summary is limited to the supplied model facts and does not infer any missing details.
Ways to run it
vLLM · Linux
Repository-specific example documented in the publisher model card; confirm dependencies and platform in the source.
vllm serve Qwen/Qwen3.5-35B-A3B --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3Transformers · See official guide
Repository-specific example documented in the publisher model card; confirm dependencies and platform in the source.
Official instructionsExplore its uses
Keep exploring
Qwen2.5 Coder · 7B
Code generation, repair, and reasoning in a compact open-weight coding model.
Mistral Small 4 · 119B
Mistral Small 4 119B A6B is an Apache-2.0 licensed hybrid model from mistralai that combines general instruction, reasoning, and Devstral-style capabilities in one model. It uses a Mixture-of-Experts architecture with 119B total parameters, 6.5B activated per token, 128 experts with 4 active, and a 256k context length. It supports multimodal input (text and image) with text output, multilingual support across dozens of languages, system prompts, function calling, JSON output, and configurable reasoning effort. The model card also mentions optional efficiency resources: an eagle head for speculative decoding and an NVFP4 checkpoint. It is tagged for vLLM and offered in safetensors format.
Qwen3-Coder-Next · 80B A3B
Qwen3-Coder-Next is an open-weight text-generation model from Qwen/Qwen3-Coder-Next, licensed under apache-2.0. It is designed specifically for coding agents and local development. The model has 80B total parameters with 3B activated parameters, a 262,144-token native context length, and a hybrid architecture combining Gated DeltaNet, Gated Attention, and Mixture-of-Experts layers. It supports only non-thinking mode and does not generate thinking blocks.