Post-Cutoff
Search For AI agents

Alibaba (Qwen)Qwen3.8 familyCurrentMultimodal modelReleased 5 Aug 2026

Qwen3.8-27B

Open-weight (Apache-2.0) dense multimodal model for local/self-hosted coding agents and vision tasks; runs on vLLM, SGLang, Transformers.

Model idqwen/qwen3.8-27b

Price$0.20 in, $2.50 out per 1M tokens

Checked25 October 2026

No published training cutoff

At least1,686events since its release on 5 August 2026

74 historic, 377 major. 100 days of news.

Alibaba (Qwen) doesn’t publish a training cutoff, so we count from its release date. The real gap is larger.

Five things it has never heard of

How to call it

5 ways to use Qwen3.8-27B, with the model id each provider expects.

Providers, model ids, endpoints and docs for Qwen3.8-27B
ProviderModel idEndpointDocs
Hugging Face huggingface.co
Hugging Face (FP8) huggingface.co
OpenRouter qwen/qwen3.8-27b openrouter.ai
OpenRouter (free tier) qwen/qwen3.8-27b:free openrouter.ai
Web app chat.qwen.ai

Pricing

Prices for Qwen3.8-27B
Input$0.20 per 1M tokens
Output$2.50 per 1M tokens

As published: “per 1M tokens, OpenRouter’s listed price for qwen/qwen3.8-27b (read from its models API 2026-10-25); its providers range from $0.02-$0.99 in and $0.50-$3.20 out, Alibaba’s own endpoint there $0.425 / $2.55; a :free variant also exists”.

Source: openrouter.ai, checked 25 October 2026.

Specs

Input
Text, images and video
Output
Text
Context window
262,144 tokens
Max output
Not published
Open weights
Yes
Licence
Apache-2.0
Training cutoff
Not published
Released
5 August 2026
Status
Current
Type
Multimodal model

What stands out

What its maker marketed at launch, and what people found later, each with its source.

  • Dense open VLM with agentic focus

    27B dense native vision-language model (images and hour-scale video) tuned for coding and long-horizon agent tasks, Apache-2.0.

    Source: huggingface.co

  • Thinking control

    Thinking on by default, can be disabled per request; reasoning_effort and preserve_thinking supported.

    Source: huggingface.co

  • Extensible to 1M context

    262,144 tokens native, extensible up to 1,000,000.

    Source: huggingface.co

Notes

Best Apache-2.0 Qwen for self-hosting; also the go-to open Qwen VL model (Qwen3-VL successor). First-party hosted API ‘coming soon’ on Qwen Cloud at time of check. No first-party price page found; OpenRouter price used (2026-10-25). ARC Prize verified (Oct 1 2026): ARC-AGI-2 42.4% at $0.45/task, ARC-AGI-1 87.5% at $0.22/task (x.com/arcprize/status/2105751611570069908).

vllm serve Qwen/Qwen3.8-27B --max-model-len 262144

Sources: https://huggingface.co/Qwen/Qwen3.8-27B · https://openrouter.ai/qwen/qwen3.8-27b