Qwen
Qwen3.8-27B
Specs
- Input price
- $0.80/M
- Cached input price
- $0.40/M
- Output price
- $4/M
- Context window
- 262.144K tokens
A dense 27B vision-language model under Apache 2.0 - agentic coding within a few points of Claude-class models, deployable on a single GPU.
Capabilities
- 27B dense stack mixing Gated DeltaNet and Gated Attention
- Native vision-language input including video, no bolt-on adapter
- Terminal-Bench 2.1 at 73.0 and SWE-bench Pro at 61.7
- OSWorld-Verified 84.3 computer use, WebArena-Verified 64.8 browser use
Best for
- Mid-tier agentic coding and office automation under Apache 2.0
- Computer-use and browser-automation agents
- Document and interface understanding without a frontier-model bill
Limitations to keep in mind
- A tier below frontier on long-horizon benchmarks (Terminal-Bench 3.0 at 30.0)
- Hosted 1M-context tier was still 'coming soon' upstream at time of writing
- Vendor-reported benchmarks on a new harness - third-party reruns pending
HIPAA-compliant hosting
Qwen3.8-27B is available under our HIPAA-compliant design-partner program, with a signed Business Associate Agreement (BAA), encryption in transit and at rest, and access controls. We are currently onboarding design partners, with general availability coming soon. Run the model behind a unified OpenRouter-style API and swap to another model with a single parameter.
Pricing in context
Open source models like Qwen3.8-27B can be served efficiently on optimized inference infrastructure, with savings passed through to you. Exact savings depend on the model and your volume, but open source inference is typically a fraction of the per-token cost of closed-source frontier models on Azure OpenAI or AWS Bedrock - without cloud egress lock-in or minimum commitments.
Design partner program
Run Qwen3.8-27B under HIPAA
Join the waitlist to be prioritized. We'll reach out with a qualification call and early access.