by datastudy.nl

The latest model launches and AI tools, explained for beginners

Subscribe
AI

Qwen 3.8 27B explained: what beginners need to know

Qwen 3.8 27B is a free, open AI model with 27 billion parameters that runs on a single GPU. It scores 61.7 on SWE-bench Pro, up from 53.5.

By Lars Cornelissen 1 min read
Abstract data visualization of Qwen 3.8 27B benchmark gains, showing rising bars from 53.5 to 61.7 percent on SWE-bench Pro and 63.9 to 84.3 percent on OSWorld-Verified against the previous Qwen3.6-27B model.
Qwen 3.8 27B benchmark improvements over Qwen3.6-27B. Source: Qwen model card on Hugging Face. Data Today benchmark.

...

Bar chart showing recommended VRAM for Qwen 3.8 27B by quantization: Q4 K M 4-bit at 17 GB, FP8 at 48 GB, and BF16 at 80 GB.
Approximate VRAM needed to run Qwen 3.8 27B at three precision levels. The 4-bit quantized version needs about 17 GB, FP8 needs a 48 GB card, and full precision BF16 needs an 80 GB card. Source: Yotta Labs. Data Today benchmark.
Grouped bar chart comparing Qwen 3.8 27B versus Qwen3.6-27B on five benchmarks. SWE-bench Pro: 61.7 vs 53.5 percent. OSWorld-Verified: 84.3 vs 63.9 percent. Terminal Bench 2.1: 73.0 vs 63.4 percent. LiveCodeBench v6: 90.3 vs 83.9 percent. CoWorkBench: 70.7 vs 61.0 percent.
Qwen 3.8 27B (blue) outperforms Qwen3.6-27B (gray) across five benchmark tasks. SWE-bench Pro rises from 53.5 to 61.7 percent, and OSWorld-Verified jumps from 63.9 to 84.3 percent. Source: Qwen model card on Hugging Face / Dell Enterprise Hub. Data Today benchmark.
#ai-for-dummies#new-model#local-ai#open-weights#quantization

Published by the Data Today newsroom · Subscribe via RSS

More from AI