Qwen Releases Qwen3.8-27B as Local Tests Reach About 65 Tokens per Second

Sign in✓ Signed in
Qwen has released Qwen3.8-27B, a 27-billion-parameter open-weight vision-language model with a native 262,144-token context window. Early local testing reports generation near 65 tokens per second on an RTX 4090, but that result used a quantized model, speculative decoding and a 65,536-token runtime configuration—not a fully occupied 262,144-token context.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.
Unlocking the full article...