pi-alibaba-token-plan 0.1.0 → 1.1.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +4 -2
- package/README.ru.md +5 -3
- package/package.json +2 -2
- package/src/catalog.js +12 -0
package/README.md
CHANGED
|
@@ -52,9 +52,11 @@ export ALIBABA_TOKEN_PLAN_REGION=sg # sg (default) or cn
|
|
|
52
52
|
pi --model alibaba-openai/qwen4-max-preview:xhigh
|
|
53
53
|
```
|
|
54
54
|
|
|
55
|
-
Thinking levels for
|
|
55
|
+
Thinking levels for `qwen4-max` are `low`, `high` and `xhigh` — only the values Alibaba documents,
|
|
56
|
+
and thinking cannot be turned off. The model has a 1M context and a 262,144-token output budget.
|
|
56
57
|
`qwen4-max-preview` has a 1M context and a 393,216-token output budget shared by thinking and the
|
|
57
|
-
answer
|
|
58
|
+
answer; its levels are `low`, `medium` and `xhigh` — ask for `:xhigh`, not `:high`. It is a
|
|
59
|
+
**preview**: Alibaba retires it after 2026-11-25.
|
|
58
60
|
|
|
59
61
|
## How the model list works
|
|
60
62
|
|
package/README.ru.md
CHANGED
|
@@ -52,9 +52,11 @@ export ALIBABA_TOKEN_PLAN_REGION=sg # sg (по умолчанию) или cn
|
|
|
52
52
|
pi --model alibaba-openai/qwen4-max-preview:xhigh
|
|
53
53
|
```
|
|
54
54
|
|
|
55
|
-
Уровни thinking у
|
|
56
|
-
|
|
57
|
-
|
|
55
|
+
Уровни thinking у `qwen4-max` — `low`, `high` и `xhigh`: только значения, которые документирует
|
|
56
|
+
Alibaba; выключить thinking нельзя. Контекст модели — 1M, бюджет вывода — 262 144 токена.
|
|
57
|
+
У `qwen4-max-preview` контекст 1M и бюджет вывода 393 216 токенов, общий для thinking и ответа;
|
|
58
|
+
уровни — `low`, `medium` и `xhigh`: просите `:xhigh`, а не `:high`. Это **превью**: Alibaba
|
|
59
|
+
отключает модель после 25.11.2026.
|
|
58
60
|
|
|
59
61
|
## Как устроен список моделей
|
|
60
62
|
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "pi-alibaba-token-plan",
|
|
3
|
-
"version": "
|
|
3
|
+
"version": "1.1.0",
|
|
4
4
|
"type": "module",
|
|
5
5
|
"description": "Alibaba Model Studio Token Plan provider for Pi: /login with region selection and a live model catalog",
|
|
6
6
|
"license": "MIT",
|
|
@@ -44,4 +44,4 @@
|
|
|
44
44
|
"qwen",
|
|
45
45
|
"token-plan"
|
|
46
46
|
]
|
|
47
|
-
}
|
|
47
|
+
}
|
package/src/catalog.js
CHANGED
|
@@ -29,6 +29,18 @@ const NON_CHAT_ID = /(^wan|image|audio|tts|asr|realtime|video|embed|rerank|ocr)/
|
|
|
29
29
|
// Models the plan serves but the catalogs do not describe yet. Specs come from
|
|
30
30
|
// https://docs.modelstudio.console.alibabacloud.com/en/model-studio/<model-id>.
|
|
31
31
|
export const VERIFIED_OVERRIDES = {
|
|
32
|
+
// Alias: qwen4.0-max. Thinking on by default (high).
|
|
33
|
+
"qwen4-max": {
|
|
34
|
+
name: "Qwen4 Max",
|
|
35
|
+
reasoning: true,
|
|
36
|
+
input: ["text", "image"],
|
|
37
|
+
contextWindow: 1_000_000,
|
|
38
|
+
// Thinking output cap.
|
|
39
|
+
maxTokens: 262_144,
|
|
40
|
+
// Docs: reasoning_effort accepts low|high|xhigh only; thinking cannot be turned off.
|
|
41
|
+
thinkingLevelMap: { off: null, minimal: null, low: "low", medium: null, high: "high", xhigh: "xhigh", max: null },
|
|
42
|
+
compat: { thinkingFormat: "qwen", supportsReasoningEffort: true, supportsStrictMode: true },
|
|
43
|
+
},
|
|
32
44
|
// Preview until 2026-11-25. Alias: qwen4.0-max-preview. Thinking on by default (xhigh).
|
|
33
45
|
"qwen4-max-preview": {
|
|
34
46
|
name: "Qwen4 Max Preview",
|