OpenAI launches ChatGPT Agent Mode priority tier — Pro and Team subscribers get expanded compute allocation, parallel tool calls, and long-horizon task persistence
OpenAI shipped a priority tier for ChatGPT Agent Mode on May 28, granting Pro ($200/month) and Team subscribers expanded compute allocation, parallel tool calls, and long-horizon task persistence across multi-hour workflows. The release positions Agent Mode against Google's Gemini Spark — launched earlier in the week — and formalizes the per-tier compute allocation pattern that frontier-lab consumer agents are converging on.
The compute-allocation tiering is the substantive piece. Free and Plus users of Agent Mode have been capped on parallel tool calls and on task-persistence horizons through the early rollout. The new priority tier removes both ceilings — Pro and Team subscribers can dispatch multiple browser sessions, code interpreter calls, and external-tool invocations in parallel, and Agent Mode can hold task state for multi-hour workflows across the user's queued tasks. The architectural commitment is the same persistent-execution pattern that Google's Spark uses, but routed through ChatGPT's existing subscription tiers rather than through a new product surface. OpenAI's product team has been explicit that the priority tier is the operational fix for Pro subscribers who hit the prior throttling on long-horizon tasks.
The competitive timing matches the consumer-agent positioning war. Google's Gemini Spark at $100/month and Anthropic's enterprise-tier Managed Agents are the two visible benchmarks; OpenAI's $200 Pro tier with Agent Mode priority is the third. The pricing-and-capability spread across the three labs is no longer principally about model capability — the underlying models are close enough that the differentiation is on execution architecture, ecosystem integration, and tier economics. Microsoft's Agent 365 enterprise rollout the same week extends the convergence into the SASE-wrapped enterprise agent traffic surface. For developer-tooling and procurement teams, the agent-platform decision is now structurally about which tier-and-ecosystem combination matches the workload.
OpenAI — ChatGPT Agent Mode priority tier launch May 28 2026 → · TechCrunch — OpenAI Agent Mode priority tier Pro Team subscribers → · The Information — ChatGPT Agent Mode compute allocation tiers →