OpenAI's GPT-5.6 Sol model with one-million-token context is now available inside the Codex CLI, according to Teknium relaying a report from Tibo (@thsottiaux). Until recently, that long-context capability reportedly worked only when routed through API keys; Tibo indicated the Codex team had just flipped a switch to enable it within Codex itself. For developers already using Codex for agent workflows, the change removes a routing split between CLI and key-based access for very large prompts or multi-file context windows. Teknium separately noted plans to update the Hermes agent framework after a related earlier error, though the packet does not specify what broke or when Hermes support will land. Treat this as a single social post relay until independent OpenAI documentation or release notes confirm rollout scope, availability tiers, and any remaining limits.
Teknium relays Tibo (@thsottiaux) reporting GPT-5.6 Sol 1M is now available in Codex. The capability previously worked only with API keys before the switch was flipped.
Key takeaway
Codex CLI users can now access GPT-5.6 Sol's 1M context without relying solely on separate API-key routing.
What happened
Teknium posted on X that Tibo (@thsottiaux) reported GPT-5.6 Sol with 1M context is now available in Codex. The post frames the change as a newly enabled capability inside Codex rather than a model that existed only outside the CLI workflow.
According to the relayed report, the 1M context option previously worked only with API keys before Codex flipped a switch to make it work in Codex. Teknium also said he would update Hermes to support the change shortly after an earlier related error.
Evidence
GPT-5.6 Sol 1M is now available in Codex per Tibo relayed by Teknium
Teknium (X) · attributed
Tibo (@thsottiaux) GPT-5.6 Sol 1M in Codex. This used to only work for API keys, but we just flipped the switch and works
The 1M context capability previously worked only with API keys before Codex enablement
Teknium (X) · attributed
Teknium relays Tibo (@thsottiaux) reporting GPT-5.6 Sol 1M is now available in Codex. The capability previously worked only with API keys before the switch was flipped.
Teknium plans to update Hermes after an earlier related error
Teknium (X) · attributed
I figured after that error earlier they might do this so will update hermes to support shortly
Why it matters
Agent builders running long-context Codex workflows may need to retest pipelines and harden framework logic where downstream tools like Hermes still lag upstream model routing changes.
Limits and uncertainties
The report comes from a single X post relay rather than official OpenAI or Codex release documentation.
The linked post excerpt is truncated and does not spell out rollout scope, pricing tiers, or remaining usage limits.
Teknium references an earlier error but the packet does not describe what failed or when Hermes support will ship.
Practical implications
Revalidate Codex CLI workflows that depend on very large context now that 1M routing may differ from API-key-only access.
Plan Hermes agent updates and defensive error handling for model-specific output or routing changes.
Avoid assuming full production readiness until independent confirmation beyond the social relay.
What to watch
Teknium's promised Hermes framework update for GPT-5.6 Sol 1M support in Codex.
Official OpenAI or Codex documentation confirming 1M context availability, limits, and access paths.
