[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]  Junyang Lin [@JustinLin610](/creator/twitter/JustinLin610) on x 27.4K followers Created: 2025-07-21 17:23:01 UTC A small update on Qwen3-235B-A22B, but a big improvement on its quality! We thought about this decision for a long time, but we believe that providing better-quality performance is more important than the unification at this moment. We are still continuing our research on hybrid thinking mode but we this time ship separate models for you! Qwen3-235B-A22B-Instruct-2507 has much more comprehensive knowledge and capabilities than the previous release, and I think you might find it a better partner for you when you use it in Qwen Chat Oh, don't forget, it natively supports 256K tokens, and with extrapolation you can extend the support to 1M tokens! Check the model card in hugging face for more detailed benchmark performance and usages! XXXXXX engagements  **Related Topics** [lin](/topic/lin) [Post Link](https://x.com/JustinLin610/status/1947346588340523222)
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
Junyang Lin @JustinLin610 on x 27.4K followers
Created: 2025-07-21 17:23:01 UTC
A small update on Qwen3-235B-A22B, but a big improvement on its quality!
We thought about this decision for a long time, but we believe that providing better-quality performance is more important than the unification at this moment. We are still continuing our research on hybrid thinking mode but we this time ship separate models for you!
Qwen3-235B-A22B-Instruct-2507 has much more comprehensive knowledge and capabilities than the previous release, and I think you might find it a better partner for you when you use it in Qwen Chat
Oh, don't forget, it natively supports 256K tokens, and with extrapolation you can extend the support to 1M tokens!
Check the model card in hugging face for more detailed benchmark performance and usages!
XXXXXX engagements
Related Topics lin