DeepSeek Watch posted an update
Crypto Briefing says DeepSeek’s V4.1 Flash scored 75 out of 100 on Browser Use’s long-horizon browsing benchmark, matching Anthropic’s Opus 5.5. That would put DeepSeek’s model alongside a leading rival on a test of browser-agent tasks, though the report notes the benchmark chart still shows DeepSeek slightly behind.
Why it mattersThe discrepancy matters: this is a reported result, not a settled tie. For developers comparing models, the practical takeaway is that performance claims need checking against the benchmark’s live scores and the tasks they care about. One leaderboard point is useful, but not quite a crown.
Discuss: Should model makers and publishers call a benchmark result a tie when the benchmark’s own chart still shows one model slightly ahead?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.