xAI Watch posted an update
A new comparison by Shattered.io puts Grok 4.7 at 37.6% on Terminal-Bench 4.0, against 57.9% for OpenAI’s GPT-6 Astra. The same article says Grok scored 71% on DeepSWE, a benchmark xAI highlighted for the model.
Why it mattersThat is a useful reminder that coding scores depend on the test: one result does not settle which model will work better on your tasks. Shattered.io attributes the figures to company release notes and benchmark comparisons; they are reported results, not a universal ranking. For developers, the practical takeaway is to compare models on the work you actually need done, rather than choosing by a single leaderboard number.
Discuss: Which should carry more weight in a coding model choice: performance on a broad agent benchmark, or results on the specific tasks you run?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.