Verified by us: this deal no longer works.
Free versions ran Sep 18 to early October; paid versions billed per token
DeepSeek V4 Flash and Qwen3.8 27B free on OpenRouter from Sep 18 to early October
The free versions of DeepSeek V4 Flash and Qwen3.8 27B ran on OpenRouter for a limited time from September 18 to early October. Both models are now billed per token, and you can switch to another free model on OpenRouter to keep testing for free.
The briefing
As of October 6, 2026, deepseek/deepseek-v4-flash:free and qwen/qwen3.8-27b:free are no longer available on OpenRouter, so these two free versions can no longer be called.
What you can use now
Both models are still available as paid versions, deepseek/deepseek-v4-flash and qwen/qwen3.8-27b, billed at the live rates shown on their OpenRouter model pages.
DeepSeek V4 Flash is DeepSeek's efficiency-focused mixture-of-experts model with about 284B total parameters and about 13B active per token, plus a 1M-token context window. Qwen3.8 27B is an open-weight dense vision-language model from Qwen, suited for coding, research, multimodal work, and long-running agent tasks.
If you want to keep testing for free, OpenRouter still has other free models you can switch to. Free models come with rate and daily request limits.
How to use it
- Sign in to OpenRouter and create an API key.
- Set your client's base URL to
https://openrouter.ai/api/v1. - Use one of the paid model IDs above, or switch to another model from the free list.
See the DeepSeek V4 Flash model page, the Qwen3.8 27B model page, and OpenRouter's free models.
