AI

Qwen3.8 27B addition in words

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 05/10/2026, 16:42:24 UTC.

Briefings

A researcher tested Qwen3.8-27B-Q4_K_M against GPT-4o on summing large numbers in words over two years. The model achieved 167/169 correct answers (99.4%) when reasoning was disabled, but performance dropped significantly when reasoning was enabled, indicating a limitation in handling complex multi-step arithmetic. The experiment was conducted on a DGX Spark using local LLMs, with results visualized as a heatmap showing success rates per number pair.

Key points

• Qwen3.8-27B-Q4_K_M achieved a 99.4% success rate (167/169) on summing large numbers in words when reasoning was disabled.
• Performance dropped significantly when reasoning was enabled, suggesting the model struggles with complex multi-step arithmetic.

Context and limitations

Results are based on a single run with reasoning enabled per number pair; the heatmap visualization is less detailed due to the reduced sample size. The experiment was conducted over two years using local hardware and may not reflect real-world conditions.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 09/10/2026, 09:26:35 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeienb53b7kw4dw2ccwl66do2jg7vnrjyqxvcgnqqa7r74x6nuxibd4