As of: 2026-11-13 13:49 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/m/qwen3-8-27b/

# Qwen3.8-27B

Alibaba (Qwen), Qwen3.8 family. Status: current. Type: multimodal model. Released: 2026-08-05. Training cutoff: not published.

## How to call it

| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Hugging Face |  |  | https://huggingface.co/Qwen/Qwen3.8-27B |
| Hugging Face (FP8) |  |  | https://huggingface.co/Qwen/Qwen3.8-27B-FP8 |
| OpenRouter | `qwen/qwen3.8-27b` |  | https://openrouter.ai/qwen/qwen3.8-27b |
| OpenRouter (free tier) | `qwen/qwen3.8-27b:free` |  | https://openrouter.ai/qwen/qwen3.8-27b |
| Web app |  |  | https://chat.qwen.ai |

## Pricing

- Input: $0.20 per 1M tokens
- Output: $2.50 per 1M tokens
- Unit 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: https://openrouter.ai/qwen/qwen3.8-27b
- Last checked: 2026-10-25

## Specs

- Input: text, image, video
- Output: text
- Context window: 262,144 tokens
- Max output: not published
- Open weights: yes
- Licence: apache-2.0

## What it has not seen

- No published training cutoff. Counted from its release (2026-08-05): at least 1,686 AI events since (377 major, 74 historic), as of 2026-11-13.
- Everything since July 2026: https://postcutoff.com/since/2026-07/

## What stands out

- 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: https://huggingface.co/Qwen/Qwen3.8-27B
- Thinking control: Thinking on by default, can be disabled per request; reasoning_effort and preserve_thinking supported. Source: https://huggingface.co/Qwen/Qwen3.8-27B
- Extensible to 1M context: 262,144 tokens native, extensible up to 1,000,000. Source: https://huggingface.co/Qwen/Qwen3.8-27B

## 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).

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

```bash
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

Model page: https://postcutoff.com/m/qwen3-8-27b/