Independent Task subagents run sequentially with GPT-5.6-sol (OpenAI)

Description

When using gpt-5.6-sol (max variant) through the OpenAI provider as the main agent, three independent Task/subagent calls requested in the same turn run sequentially. Each task completes successfully; none is stuck or fails.

Actual: Subagent A finishes before B starts, and B finishes before C starts. Total duration is approximately A + B + C.

Expected: Independent subagents should start concurrently, with total duration close to max(A, B, C).

There is no custom Task/Subagent configuration. I reviewed #11865 and #37315; this report differs because no subagent is stuck or failing. This is a corrected resubmission of #46451.

Plugins

vision-bridge, read-pdf, anchored-standard, keyword-skills, append-question. None modifies Task/Subagent scheduling.

OpenCode version

1.18.25

Steps to reproduce

1. Configure gpt-5.6-sol (max) from the OpenAI provider as the main model.

2. Start a session and give the main agent three independent tasks with no dependencies.

3. Use this prompt:

   Delegate these three independent tasks to three separate subagents. Start all three in the same turn and run them concurrently. Do not wait for one task to finish before starting another.

4. Observe that B starts only after A finishes, and C starts only after B finishes.

Screenshot and/or share link

Previous auto-closed report: https://github.com/anomalyco/opencode/issues/46451

Operating System

Windows 11 Pro for Workstations 10.0.26200 (Build 26200)

Terminal

OpenCode TUI in PowerShell 7 (pwsh)