# OpenAI models > Pick an OpenAI model ref, including GPT-6.1 Sol, GPT-6 Astra, Sol, Luna, and the GPT-5.6 tiers - 网址：https://funcoding.ai/agents/openclaw/providers/openai/models/ - 来源：OpenClaw 官方文档原文（英文），MIT 许可，同步于 2026-10-11 - 官方原文：https://docs.openclaw.ai/zh-CN/providers/openai/models --- ## Quick choice | Goal | Use | Notes | | ------------------------------------------------- | ------------------------------------------------------------------ | ------------------------------------------------------------------- | | ChatGPT/Codex subscription, native Codex runtime | `openai/gpt-6-astra` | Fresh subscription setup; sign in with Codex auth. | | Direct API-key billing for agent turns | `openai/gpt-6-astra` plus an ordered API-key auth profile | Fresh API-key setup uses Astra. | | GPT-6.1 Sol | `openai/gpt-6.1-sol` | Select explicitly; reasoning cannot be disabled. | | GPT-6 Sol | `openai/gpt-6-sol` | Select explicitly; account access can differ between auth routes. | | Lower-cost GPT-6 Luna | `openai/gpt-6-luna` | Select explicitly; check the account catalog for availability. | | Choose an exact GPT-5.6 tier | `openai/gpt-5.6-sol`, `-terra`, or `-luna` | Check `models list` for the tiers available to this account. | | Account without GPT-5.6 access | `openai/gpt-5.5` | Explicit recovery choice; OpenClaw does not silently downgrade. | | Direct API-key billing, explicit OpenClaw runtime | `openai/gpt-5.6` plus provider/model `agentRuntime.id: "openclaw"` | Select a normal `openai` API-key profile. | | Latest ChatGPT Instant model alias | `openai/chat-latest` | Direct API-key only; moving alias, not the stable default. | | Image generation or editing | `openai/gpt-image-2` | Works with `OPENAI_API_KEY` or Codex OAuth. | | Transparent-background images | `openai/gpt-image-1.5` | Set `outputFormat` to `png` or `webp` and `background=transparent`. | Isolated and utility completions on the Codex runtime use the same account routes as the model picker. A model listed by your ChatGPT account does not require an OpenAI API key for these completions. An explicit auth profile or utility-model override still controls the request; OpenClaw does not replace it with another account. ### Retired subscription model references GPT-5.4 and GPT-5.4 Mini are retired from the ChatGPT-account Codex route. Run `openclaw doctor --fix` to replace persisted subscription references with their documented successors: `openai/gpt-5.6-terra` and `openai/gpt-5.6-luna`, respectively. This includes defaults, per-agent model selections, automation overrides, and unlocked session overrides whose selected route is known. The Platform API-key route is unaffected. Doctor retains pinned overrides when their successor is outside the agent's model policy, or when clearing an override would keep the same retired model and account. Doctor also retains the original reference and warns when its declared successor is retired or definitively unsupported on the selected account route; unknown availability and temporary cooldowns do not block migration. It reports the model or policy change needed, along with unresolved or conflicting account routes. Review the repair output, restart the Gateway, and re-enable any automation that was disabled after repeated failures. ## Models your API key lists With an OpenAI API key, `openclaw models list --provider openai` shows every chat model your account's `/v1/models` lists that runs on the Responses API, including models newer than your OpenClaw release. Models without a catalog row use conservative defaults: text input, a 128k context window, and unknown cost. Embedding, image, video, audio, realtime, speech, transcription, moderation, and legacy completions models stay hidden, as do models past their OpenAI shutdown date. Search models stay hidden because they run only on Chat Completions; live, cyber, experiment, and alpha IDs stay hidden because Responses rejects them. GPT-3.5, GPT-4 (including GPT-4o and GPT-4.1), o1, o3, and o4 models stay hidden from the list because they fail on the default Codex runtime. You can still select one by its reference, for example as `agents.defaults.model.primary` or a session model; set `agents.defaults.models["openai/"].agentRuntime.id` to `"openclaw"` so it runs. o1, o3-mini, GPT-3.5, GPT-4, GPT-4 Turbo, and GPT-4.1 nano also reject OpenAI's hosted web search, which that runtime adds by default: set `tools.web.search.provider` to a managed provider such as `brave`, or `tools.web.search.enabled: false`, before using them (see [Native OpenAI web search](https://funcoding.ai/agents/openclaw/tools/web/#native-openai-web-search)). ## Daybreak Blue and Red Accounts provisioned for Daybreak can select `openai/gpt-daybreak-blue-latest` or `openai/gpt-daybreak-red-latest` with an OpenAI API-key profile. Both aliases use the Responses API and expose `low`, `medium`, `high`, `xhigh`, and `max` on the OpenClaw runtime. `/think ultra` remains a separate orchestration mode; it uses the highest supported native effort rather than sending `ultra` to the API. On the OpenClaw runtime, Daybreak Blue supports Standard and Fast processing; Daybreak Red uses Standard. Neither alias supports Ultrafast. The Control UI disables unavailable speed choices, and requests apply the same limits: saved Ultrafast preferences use Fast on Blue and Standard on Red. Explicit low-level `serviceTier` / `service_tier` overrides remain operator-controlled and may be rejected by the API. Daybreak aliases can resolve to different snapshots as access programs evolve. OpenClaw preserves the requested alias instead of replacing it with a snapshot. Blue keeps reasoning enabled; Red also supports `/think off`. Explicit configured effort capabilities remain authoritative, including narrower or empty lists. Alias cost estimates are unknown, not a claim that Daybreak requests are free; check the resolved model and current OpenAI pricing. Successful account discovery remains authoritative and this support does not grant Daybreak access. Native Codex continues to use its account catalog and native effort controls. See the [Daybreak guide](https://developers.openai.com/api/docs/guides/daybreak) and the [Blue](https://developers.openai.com/api/docs/models/gpt-daybreak-blue-latest) and [Red](https://developers.openai.com/api/docs/models/gpt-daybreak-red-latest) model references. ## GPT-6 Astra Select `openai/gpt-6-astra` with an OpenAI API-key profile or a ChatGPT/Codex subscription that has access to Astra. Access is rolling out; a successful account catalog remains authoritative, so adding model support does not grant access to an account that has not received it. If ChatGPT/Codex catalog discovery is unavailable, the offline fallback list omits Astra until account discovery succeeds. ```bash openclaw models set openai/gpt-6-astra ``` Astra uses the Responses API for agent tool calls. It supports text and image input, a 1,050,000-token context window, and up to 128,000 output tokens. OpenClaw retains its ordinary 272,000-token active input budget by default. The supported reasoning efforts are `low`, `medium`, `high`, `xhigh`, and `max`. OpenClaw defaults Astra to `medium` on both the OpenClaw and Codex runtimes when the account supports that effort. The OpenAI provider owns this default, so model selection, Control UI, and Codex turn requests share it. Explicit agent, model, global, and session thinking settings still take precedence; switching models does not clear an existing `high` override. Use `/think default` to clear a session override. An existing `minimal` setting maps to `low`. Astra cannot disable reasoning; `off` never sends the unsupported `none` effort. Temperature and `top_p` are not sent. These defaults also apply to configured Astra model entries without explicit reasoning or temperature compatibility metadata. Azure Responses deployments continue to use their configured capabilities. `/think ultra` is also available on the OpenClaw and Codex runtimes. Ultra enables proactive sub-agent orchestration; it is not a raw Responses API effort. OpenClaw uses `max`, while native Codex selects Astra's model-defined effort (`xhigh`). Standard pricing per million tokens is $10 input, $1 cache reads, $12.50 cache writes, and $50 output. Requests above 272K input tokens have higher rates. See the [Astra model reference](https://developers.openai.com/api/docs/models/gpt-6-astra) and [migration guide](https://developers.openai.com/api/docs/guides/latest-model?model=gpt-6-astra). ### Async tools, steering, and reasoning changes Use an OpenAI Platform API-key profile and the built-in OpenClaw runtime for these Astra capabilities. They require the official `https://api.openai.com/v1` Responses endpoint. Configure the existing model settings: ```json5 { agents: { defaults: { models: { "openai/gpt-6-astra": { agentRuntime: { id: "openclaw" }, params: { transport: "auto", responsesServerCompaction: false, }, }, }, }, }, } ``` - **Async function calls:** Astra can continue reasoning while OpenClaw runs a direct function tool. OpenClaw sends the completed result in the next model request after the active response finishes. This applies to direct tools; code-mode tools retain their existing execution flow. `sessions_yield` stays synchronous: it is how the model waits, so the response pauses there and the next request delivers the earlier async results. - **Mid-turn steering:** [Steering messages](https://funcoding.ai/agents/openclaw/concepts/queue/#queue-modes) can reach Astra while it is reasoning, using the active session's cached WebSocket. Use `auto` or `websocket-cached`; SSE keeps ordinary queued steering at the next available runtime boundary. Each live batch owns one response; later messages can steer its successor. Context or payload hooks that rewrite the active request's prefix keep ordinary queued delivery. - **Reasoning changes without rebuilding the cached prefix:** Change the [thinking level](https://funcoding.ai/agents/openclaw/tools/thinking/), for example with `/think high`, before the next user turn. OpenClaw preserves the original request-level effort and places a `configuration_update` at the new turn. This optimization works across matching session history over SSE or cached WebSockets. Automatic steering continuations keep their inherited settings. If steering waits for a tool result or approval, the explicit continuation uses current request settings, including output limits and reasoning settings, without repeating accepted steering. Earlier `configuration_update` items retain their effect; a changed request-level effort does not replace those controls. When accepted steering waits for a tool result or approval and its history contains effort controls, finish that input with a compatible Astra model and mode before switching. The example disables automatic server compaction because OpenAI cannot combine it with configuration updates. Cache-preserving effort changes also exclude automatic truncation, pro mode, and API multi-agent mode. The original effort and admitted controls survive transport expiry and Gateway restarts in saved provider replay metadata. Matching history replays the same prefix without extending socket or provider cache lifetimes. Rewritten or compacted history, incompatible settings, or a changed model, route, session, or auth profile starts a fresh request using the selected effort. Older transcripts without this metadata also start fresh after transport expiry. The native [Codex harness](https://funcoding.ai/agents/openclaw/plugins/codex-harness/) owns its own Responses loop; these built-in-runtime capabilities do not imply native Codex support. ## GPT-6.1 Sol Select `openai/gpt-6.1-sol` with an OpenAI API-key profile or a ChatGPT/Codex subscription that exposes it: ```bash openclaw models set openai/gpt-6.1-sol ``` It uses the Responses API for tool calls, accepts text and images, and supports a 1,050,000-token context window with up to 128,000 output tokens. OpenClaw keeps its 272,000-token active input budget by default; native Codex follows the selected account's advertised limits. Existing selections and the Astra setup default are unchanged. A successful account catalog remains authoritative; subscription offline hints do not imply access. Reasoning defaults to `medium` and supports `low`, `medium`, `high`, `xhigh`, and `max`. Unlike GPT-6 Sol, GPT-6.1 Sol cannot disable reasoning: `off` does not send `none`, and `minimal` maps to `low`. OpenClaw offers `/think ultra` through its existing orchestration mode; native Codex uses the account's advertised efforts. Standard API prices per million tokens are $2 input, $0.10 cache reads, $2.50 cache writes, and $10 output. Above 272K input tokens, input and cache rates double and output costs 1.5 times the standard rate for the full request. See the [GPT-6.1 Sol model reference](https://developers.openai.com/api/docs/models/gpt-6.1-sol) for current capabilities, pricing, and regional availability. ## GPT-6 Sol and Luna Select `openai/gpt-6-sol` or `openai/gpt-6-luna` with an OpenAI API-key profile or a ChatGPT/Codex subscription that exposes the model. Check the selected account's catalog, then choose the model: ```bash openclaw models list --provider openai openclaw models set openai/gpt-6-sol ``` Use `openclaw models set openai/gpt-6-luna` to select Luna. API organization and Codex workspace access can differ. A successful account catalog is authoritative; OpenClaw does not add subscription access or silently substitute another model. When subscription discovery is unavailable, the offline fallback list omits both models. Existing model selections stay unchanged, and fresh OpenAI setup continues to use Astra. Both models use the Responses API for agent tool calls and support text and image input, a 1,050,000-token context window, and up to 128,000 output tokens. OpenClaw defaults to a 272,000-token active input budget. Native Codex follows the selected account's advertised context limits. The API reasoning efforts are `none`, `low`, `medium`, `high`, `xhigh`, and `max`. OpenClaw defaults to `medium` when available; existing explicit thinking settings still take precedence. `/think off` selects `none`, and `/think default` clears a session override. Native Codex uses the effort levels reported by the selected account. OpenClaw's `/think ultra` mode uses `max`; native Codex offers Ultra only when the account advertises it. Standard API pricing per million tokens: | Model | Input | Cache reads | Cache writes | Output | | ---------- | ----- | ----------- | ------------ | ------ | | GPT-6 Sol | $2 | $0.20 | $2.50 | $10 | | GPT-6 Luna | $0.10 | $0.01 | $0.125 | $0.50 | See the [GPT-6 Sol model reference](https://developers.openai.com/api/docs/models/gpt-6-sol) and [GPT-6 Luna model reference](https://developers.openai.com/api/docs/models/gpt-6-luna) for current capabilities and pricing. ## GPT-5.6 limited preview OpenClaw recognizes the exact `openai/gpt-5.6-sol`, `openai/gpt-5.6-terra`, and `openai/gpt-5.6-luna` model ids. All three expose `xhigh` and `max` reasoning in the current catalog. OpenAI describes Sol as the flagship tier, Terra as the balanced tier, and Luna as the fast, lower-cost tier. See the [GPT-5.6 launch announcement](https://openai.com/index/previewing-gpt-5-6-sol/) and [access guide](https://help.openai.com/en/articles/20001325-a-preview-of-gpt-5-6-sol-terra-and-luna). OpenAI's [GPT-5.6 Sol model page](https://developers.openai.com/api/docs/models/gpt-5.6-sol) documents the bare `openai/gpt-5.6` id as a supported alias for Sol. Fresh API-key and ChatGPT/Codex OAuth setup use `openai/gpt-6-astra`. Existing GPT-5.6 selections retain their canonical Sol identity. Run `openclaw doctor --fix` to rewrite persisted bare OpenAI refs to that canonical identity. The native Codex catalog can show the exact Sol, Terra, and Luna ids depending on workspace access. Check the current account with: ```bash openclaw models list --provider openai ``` API organization and Codex workspace access can differ. If GPT-5.6 is not available, select GPT-5.5 explicitly: ```bash openclaw models set openai/gpt-5.5 ``` OpenClaw surfaces the upstream access error and does not silently replace a GPT-5.6 selection with GPT-5.5.
Eligible exact official HTTPS routes may select the bundled Codex app-server plugin when runtime policy is unset or `auto`; authored Completions routes, custom endpoints, and request-transport overrides remain on OpenClaw. Plaintext official HTTP endpoints are rejected. Explicit provider/model runtime config remains authoritative. Run `openclaw doctor --fix` to repair stale legacy Codex model refs, `codex-cli/*` refs, or old runtime session pins that were not set by explicit runtime config.