Latest Content
Qwen3.5 0.8B icon

Qwen3.5 0.8B

NVIDIA
Qwen3.5 0.8B is a compact multimodal language model combining a causal transformer with a vision encoder for efficient cross-modal reasoning. It features a 0.8B parameter architecture with 24 layers, 1024 hidden size, 8 attention heads (2 KV heads), and integrates Gated DeltaNet linear attention with periodic full attention layers. The model supports a 262K token context window and uses multi-token prediction (MTP). It is designed for efficient multimodal understanding, strong reasoning, and broad multilingual support across 200+ languages with low-latency inference.
TypeVision-Language Model
CapabilitiesText Generation, Instruction Following, Reasoning, Mathematical Reasoning+6 more
Release Date02 March, 2026
Links
Blog|HF Model Card
LicenseApache 2.0

Inference Instructions

Deploy and run this model on NVIDIA B200 GPUs using the command below. Copy the command to get started with inference.

CONSOLE
docker run --gpus all 
 --shm-size 128g 
 -p 8000:8000 
 -v ~/.cache/huggingface:/root/.cache/huggingface 
 -e HF_TOKEN='YOUR_HF_TOKEN' 
 --ipc=host 
 lmsysorg/sglang:v0.5.9 
 python3 -m sglang.launch_server 
 --model-path Qwen/Qwen3.5-0.8B 
 --host 0.0.0.0 
 --port 8000 
 --max-prefill-tokens 65536 
 --max-running-requests 1024 
 --tp 4 
 --tool-call-parser qwen3_coder 
 --reasoning-parser qwen3 
 --mem-fraction-static 0.95 
 --trust-remote-code
Note

Multimodal serving is constrained to a maximum of tp=4; for higher parallelism (e.g., tp=8) or text-only inference, the --language-only flag must be enabled.

Model Benchmarks

Each model was tested with a fixed input size and total token volume while increasing concurrency to measure serving performance under load.

ITL vs Concurrency

Time to First Token

Throughput Scaling

Total Tokens/sec vs Avg TTFT

Vultr Cloud GPU

NVIDIA HGX B200

Deploy NVIDIA B200 on Vultr Cloud GPU

Deploy NVIDIA B200 instancesContact Sales