You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
- adaptive → provider-managed adaptive reasoning budget (supported for Anthropic Claude 4.6 model family)
18
+
- xhigh → “ultrathink+” (GPT-5.2 + Codex models and Anthropic Claude Opus 4.7)
19
+
- adaptive → provider-managed adaptive reasoning budget (supported for Anthropic Claude 4.6 and Opus 4.7)
20
20
-`x-high`, `x_high`, `extra-high`, `extra high`, and `extra_high` map to `xhigh`.
21
21
-`highest`, `max` map to `high`.
22
22
- Provider notes:
23
-
- Anthropic Claude 4.6 models default to `adaptive` when no explicit thinking level is set.
23
+
- Anthropic Claude 4.6 and Opus 4.7 models default to `adaptive` when no explicit thinking level is set.
24
+
- Anthropic Claude Opus 4.7 maps `/think xhigh` to `output_config.effort: "xhigh"`.
24
25
- MiniMax (`minimax/*`) on the Anthropic-compatible streaming path defaults to `thinking: { type: "disabled" }` unless you explicitly set thinking in model params or request params. This avoids leaked `reasoning_content` deltas from MiniMax's non-native Anthropic stream format.
25
26
- Z.AI (`zai/*`) only supports binary thinking (`on`/`off`). Any non-`off` level is treated as `on` (mapped to `low`).
26
27
- Moonshot (`moonshot/*`) maps `/think off` to `thinking: { type: "disabled" }` and any non-`off` level to `thinking: { type: "enabled" }`. When thinking is enabled, Moonshot only accepts `tool_choice``auto|none`; OpenClaw normalizes incompatible values to `auto`.
@@ -31,7 +32,7 @@ title: "Thinking Levels"
31
32
2. Session override (set by sending a directive-only message).
32
33
3. Per-agent default (`agents.list[].thinkingDefault` in config).
33
34
4. Global default (`agents.defaults.thinkingDefault` in config).
34
-
5. Fallback: `adaptive` for Anthropic Claude 4.6 models, `low` for other reasoning-capable models, `off` otherwise.
35
+
5. Fallback: `adaptive` for Anthropic Claude 4.6 and Opus 4.7 models, `low` for other reasoning-capable models, `off` otherwise.
35
36
36
37
## Setting a session default
37
38
@@ -104,8 +105,9 @@ title: "Thinking Levels"
104
105
105
106
- The web chat thinking selector mirrors the session's stored level from the inbound session store/config when the page loads.
106
107
- Picking another level writes the session override immediately via `sessions.patch`; it does not wait for the next send and it is not a one-shot `thinkingOnce` override.
107
-
- The first option is always `Default (<resolved level>)`, where the resolved default comes from the active session model: `adaptive` for Claude 4.6 on Anthropic/Bedrock, `low` for other reasoning-capable models, `off` otherwise.
108
+
- The first option is always `Default (<resolved level>)`, where the resolved default comes from the active session model: `adaptive` for Claude 4.6 and Opus 4.7 on Anthropic, `low` for other reasoning-capable models, `off` otherwise.
108
109
- The picker stays provider-aware:
109
110
- most providers show `off | minimal | low | medium | high | adaptive`
111
+
- Anthropic Claude Opus 4.7 shows `off | minimal | low | medium | high | xhigh | adaptive`
110
112
- Z.AI shows binary `off | on`
111
113
-`/think:<level>` still works and updates the same stored session level, so chat directives and the picker stay in sync.
0 commit comments