The Qwen3.5 27B model did almost the same as Sonnet 4.5 in this[1] reasoning benchmark, results here[2]. Obviously there's more to a model than that but it's a

The Qwen3.5 27B model did almost the same as Sonnet 4.5 in this[1] reasoning benchmark, results here[2]. Obviously there's more to a model than that but it's a data point. [1]: https://github.com/fairydreaming/lineage-bench https://github.com/fairydreaming/lineage-bench [2]: https://github.com/fairydreaming/lineage-bench-results/tree/main/lineage-8_64_128_192#results https://github.com/fairydreaming/lineage-bench-results/tree/...