Supported Models

Transformer-based model architectures supported by the Furiosa SDK, including decoder-only, pooling, and vision-language models.

FuriosaAI's software stack supports a wide range of Transformer-based models available on the Hugging Face Hub. Below is a list of model architectures currently supported by the Furiosa SDK. If your model is based on any of these architectures, you can use the Furiosa SDK to compile, quantize, and run the model efficiently on Furiosa's NPUs.

TIP

Pre-compiled models are available in Hugging Face Hub 🤗 - FuriosaAI organization. You can download and run them quickly with Furiosa-LLM.

Decoder-only Models (Text Generation)

Model NameArchitectureExample Hugging Face Models
DeepSeek R1LlamaForCausalLMdeepseek-ai/DeepSeek-R1-Distill-Llama-8B, deepseek-ai/DeepSeek-R1-Distill-Llama-70B
EXAONE 4.0Exaone4ForCausalLMLGAI-EXAONE/EXAONE-4.0-32B, LGAI-EXAONE/EXAONE-4.0-32B-FP8
Llama 3.1, Llama 3.3LlamaForCausalLMmeta-llama/Llama-3.1-8B-Instruct, meta-llama/Llama-3.1-70B-Instruct, meta-llama/Llama-3.3-70B-Instruct
Solar 1.0LlamaForCausalLMupstage/SOLAR-10.7B-v1.0, upstage/SOLAR-10.7B-Instruct-v1.0
Qwen 2, Qwen 2.5Qwen2ForCausalLMQwen/Qwen2.5-Coder-32B-Instruct, Qwen/Qwen2-32B
Qwen 3Qwen3ForCausalLMQwen/Qwen3-0.5B-Instruct, Qwen/Qwen3-32B-Instruct

Pooling Models

Model NameArchitectureTaskExample Hugging Face Models
Qwen 3 EmbeddingQwen3ModelEmbeddingQwen/Qwen3-Embedding-4B, Qwen/Qwen3-Embedding-8B
Qwen 3 RerankerQwen3ForSequenceClassificationRerankingQwen/Qwen3-Reranker-4B, Qwen/Qwen3-Reranker-8B

Vision-Language Models (Multimodal)

See Vision-Language Models for a guide on launching a VL server and sending image inputs over the OpenAI-compatible Chat Completions API.

Model NameArchitectureModalitiesExample Hugging Face Models
Qwen 3 VLQwen3VLForConditionalGenerationText, ImageQwen/Qwen3-VL-32B-Instruct, Qwen/Qwen3-VL-2B-Instruct-FP8

Planned Models for Future Releases

(The order of the models listed below indicate the priority of support.)

  • GPT-OSS
  • K-EXAONE
  • Qwen3 MoE

On this page