Models and Reasoning

Current Model Baseline (2026-10-23)

The models page recommends Astra, GPT-6.1 Sol, and GPT-6 Luna. GPT-6.1 Sol, GPT-6 Sol, and GPT-6 Luna are available in Work and Codex in ChatGPT, not in Chat. The GPT-6.1 Sol launch rollout covers Plus, Pro, Business, Enterprise, and Edu; Free and Go are not included. Availability depends on the rollout, sign-in method, and client; Enterprise and Edu administrators must enable the new models first, and selecting one does not grant workspace access.

GPT-6.1 Sol Specifications

Cached input costs 5% of the uncached input rate, lower than GPT-6 Sol's 10%. Requests over 272K input tokens are billed at 2x input and cache rates and 1.5x output for the whole request; Fast costs 2x standard, and Batch and Flex cost 50%. US and EU data residency are supported, but Fast mode is unavailable with EU data residency. The official docs set no specific starting effort for GPT-6.1 Sol: start from the client default and adjust.

GPT-6 Sol and Luna Specifications

For both models, requests over 272K input tokens are billed at 2x input and cache rates and 1.5x output for the whole request. Batch and Flex cost 50% of standard; fast mode costs 2x. Reasoning levels don't map exactly across model generations, so retry a familiar task at a lower level first.

GPT-6 Astra Specifications

Operating Rules

1. Use Astra plus high~max for the hardest and longest work: large-context migrations, multistep research, and tasks mixing computer use with code.
2. Use Astra or GPT-6.1 Sol plus high~max and human review for ambiguous or high-blast-radius work.
3. Use GPT-6.1 Sol for everyday engineering, starting from the client's default effort; raise effort or escalate to Astra when failure cost rises.
4. Start GPT-6 Luna at high for explicit, repeatable tasks and lower it once results hold.
5. Use Ultra when the goal benefits from automatic delegation, after bounding write scopes and budget.
6. Use Max for deeper single-task reasoning; do not treat it as an alias for Ultra orchestration.

Astra's API price is 5x GPT-6.1 Sol and 100x Luna. With GPT-6.1 Sol as the bundled default from 0.159.1, sessions with an unset model cost less, but run representative task evals to confirm that work previously run on Astra keeps its quality, and keep repetitive work on Luna.

Commands and Configuration

codex --model gpt-6-astra
codex --model gpt-6.1-sol  # bundled default from 0.159.1
codex exec -m gpt-6.1-sol "Review the current changes"
codex --model gpt-6-sol    # 0.156.1+
codex --model gpt-6-luna

# Previous generation during the rollout (regression baseline)
codex --model gpt-5.6-sol

Authentication, account, and provider rollout can change the exact IDs exposed to a workspace. Use the /model picker as the final availability check.

model = "gpt-6-astra"
model_reasoning_effort = "high"

# Balanced everyday default
# model = "gpt-6.1-sol"
# model_reasoning_effort = "medium"

# Optional dedicated review model
# review_model = "gpt-6-astra"

Provider and Cloud Boundaries

• Codex CLI 0.148.0 adds a built-in amazon-bedrock provider configured with AWS profile and region. GPT-6 Astra appears in the Bedrock Mantle and Runtime pickers from 0.153.3, and GPT-6 Sol and Luna from 0.157.0; from 0.159.1, GPT-6.1 Sol is the default in both catalogs. 0.161.0 adds multi-agent V2 and Ultra reasoning on compatible Bedrock models and lets Bedrock Mantle accept AWS GovCloud regions; the release notes do not list the compatible models, so check /model in the Bedrock account. From 0.160.0, an explicit provider model catalog no longer includes unsupported bundled models. Since 0.155.0, Bedrock can obtain AWS credentials from configured commands, with caching and expiry-based refresh.
• Do not assume direct OpenAI, Bedrock, and custom providers expose identical IDs or reasoning levels; record each provider's /model inventory and eval results.
• A local CLI/App/IDE model change does not change the fixed default model of a Codex Cloud task.

Migration Checklist

• Upgrade to CLI 0.153.0 or later and confirm with codex --version. Pin model in config.toml if you do not want the 0.153.4 default change to apply.
• Review /permissions, sandboxing, and auto-approval scope before adopting Astra, given its Critical cybersecurity classification.
• Upgrade to CLI 0.156.1 or later (0.157.0 for Bedrock) and confirm GPT-6 Sol and Luna in the picker.
• On CLI 0.159.1 or later, confirm with /status that sessions with an unset model run on GPT-6.1 Sol. Pin model = "gpt-6-astra" to keep Astra as the default.
• Before 2026-10-14, replace gpt-5.5 on the ChatGPT-sign-in path with GPT-6 Sol (Luna on Free and Go).
• Move gpt-5.6-* defaults to GPT-6 Sol and Luna; keep the previous generation only as a regression baseline.
• Clean up remaining gpt-5.4, gpt-5.4-mini, gpt-5.3-codex-spark, gpt-5.2, and gpt-5.3-codex references.
• In Enterprise and Edu, confirm an administrator has enabled GPT-6.1 Sol, GPT-6 Sol, and Luna in workspace model availability.
• Confirm the selected model and reasoning effort with /model and /status.
• Compare Astra, GPT-6.1 Sol, and Luna for quality, latency, and credit use on representative team tasks.
• Keep Cloud and API-key provider policies separate.

Sources

• Models: https://learn.chatgpt.com/docs/models
• GPT-6 Astra (API): https://developers.openai.com/api/docs/models/gpt-6-astra
• GPT-6.1 Sol (API): https://developers.openai.com/api/docs/models/gpt-6.1-sol
• GPT-6 Sol (API): https://developers.openai.com/api/docs/models/gpt-6-sol
• GPT-6 Luna (API): https://developers.openai.com/api/docs/models/gpt-6-luna
• GPT-6 Astra announcement: https://openai.com/index/gpt-6-astra/
• Codex CLI: https://learn.chatgpt.com/docs/codex/cli
• Slash commands: https://learn.chatgpt.com/docs/developer-commands?surface=cli
• Config reference: https://learn.chatgpt.com/docs/config-file/config-reference
• Changelog: https://learn.chatgpt.com/docs/changelog