**OpenAI’s recent GPT-6 series releases, including GPT-6.1 Sol on September 29, 2026, have emphasized agentic coding, computer use, and cost-efficient performance close to GPT-6 Astra, yet HLE leaderboards as of October 2 remain led by Anthropic’s Claude Fable 5.1 and Opus 5 variants at 64.5–65%.** GPT-6 Astra posted an 82.9% score on the refined HLE-Diamond subset (with tools) in late September, but main HLE results for GPT-5.4/5.6 Sol models trail the frontier, and GPT-6.1 Sol announcements highlighted DeepSWE and workflow gains without new HLE figures. The benchmark’s expert-curated questions continue to favor models demonstrating sustained reasoning depth, where Claude variants currently hold the edge. Upcoming catalysts include any OpenAI or Anthropic system-card updates citing HLE, potential tool-augmented re-evaluations, or the next Sol point release, all of which could shift trader views on whether a 6.1+ Sol model will mark its public benchmark debut on this frontier test.