Anthropic leads trader sentiment at 72.5% implied probability due to its September 2026 launches of Claude Opus 5.5 and Sonnet 5.5, which deliver frontier-level agentic coding, tool use, and long-horizon task performance at 30-40% lower cost than prior flagships, alongside the production rollout of Claude Managed Agents and a NVIDIA partnership for the Open Agent Safety Platform emphasizing secure sandboxing and execution controls. These updates build on computer-use capabilities, multi-agent orchestration, and enterprise integrations such as Barclays' scaling of Claude Code, with Anthropic models topping recent agent benchmarks like BenchAlign and Terminal-Bench. OpenAI, Google, and others trail at 14% or below amid their own agent releases, but face tighter competition on cost, security, and demonstrated autonomous workflows through December.