Vals AI Says GPT 6 Astra Leads Code Migration Benchmark
The startup's benchmark ranks the model first in accuracy for rewriting code in new languages.
TLDR
Vals AI announced that GPT 6 Astra achieved the top score on its Code Migration benchmark. The model reached 68 percent accuracy, ten points ahead of the next entry. It also runs two to four times faster than comparable models. The result comes from the company's public evaluation of large language models on real-world tasks. A scatter plot shared in the post displays accuracy against cost for various systems on the code rewriting task. Vals AI provides these benchmarks for finance, law, and similar domains.
Combined views
32.1K
1 Source, first seen 27d ago