Skip to content

Qwen3.5 4B Qwen3.8 27B Draft

nvidia/Qwen3.5-4B-Qwen3.8-27B-Draft
Open Source · chat · open-weights
Open Alert me on changes
Context
1M
Max output
—
Weights
Open
API $/1M
$0.04 / $0.90
Modalities
text
Released
11 Nov 2026
Download image Share on X Share on LinkedIn
AI summary
● machine-written

Qwen3.8 27B Draft: Open-weight vision-language model with 1M context

Qwen3.8 27B is an open-weight dense vision-language model released August 14, 2026, designed for coding, professional workflows, research, and multimodal tasks. The model supports a 1.0M token context window and is available through 25+ providers with pricing starting at $0.035 per million input tokens. It includes flexible thinking capabilities that can be enabled or disabled.

What's new
  • 1.0M token context window
  • Vision-language capabilities with flexible thinking
  • Available from 25+ providers including Darkbloom, Modal, DekaLLM
  • Input prices range from $0.035 to $0.99 per million tokens across providers
  • Benchmarks: 81.8 on GPQA (63rd percentile), Intelligence 10.0 (7th percentile)
Best for
Coding and professional workflowsMultimodal interaction and researchLong-running agent tasksVision-language applications
Sources

Source: https://huggingface.co/nvidia/Qwen3.5-4B-Qwen3.8-27B-Draft