User reports Qwen 3.8 Max Preview more token-efficient in input costs
A Reddit user reports Qwen 3.8 Max Preview is "much more token efficient" in input costs when tested with the same per-turn prompt/system instructions against Minimax M3, 5.6 Sol X High, and 5.6 Luna High. The post is a follow-up correcting an earlier mention of the model as 3.7 Max Preview.
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Smaller, faster, safer: running Kimi and GLM at scale
- Prometheus automates bug reproduction and patch generation
- China's Next AI Race Goes Beyond Chatbots
- Detect Energy Theft Faster with Genie
- Tool runs hundreds of large models on a single GPU, cold starts under 2s