Skip to content

A local-first multi-agent app development platform, built on PenguinHarness

fix(core): move the DeepSeek V4 Flash rows to the price effective 2026-09-10 (#649)

Yaowei Zheng340193ca1887
diff --git a/changelog/unreleased/2026-09-08-deepseek-flash-pricing.md b/changelog/unreleased/2026-09-08-deepseek-flash-pricing.md
new file mode 100644
index 00000000..7fe1d570
--- /dev/null
+++ b/changelog/unreleased/2026-09-08-deepseek-flash-pricing.md
@@ -0,0 +1,59 @@
+# The DeepSeek V4 Flash price effective 2026-09-10, and a pre-registered V4.1 Flash row
+
+- **Date:** 2026-09-08
+- **Type:** fix
+- **Scope:** `core`, `cli`, `docs`, `skills`
+- **PR:** [#649](https://github.com/Prism-Shadow/penguin-harness/pull/649)
+
+[中文版](2026-09-08-deepseek-flash-pricing.zh.md)
+
+DeepSeek adjusted the V4 Flash series price, effective 2026-09-10 12:00 Beijing. The catalog's two
+direct V4 Flash rows were re-read on 2026-09-08 and now record the peak tier CNY 0.04 / 2 / 8 per
+million tokens (cache hit / cache miss / output), whose off-peak half — 0.02 / 1 / 4 — is applied
+from the schedule each row already declares. `deepseek-v4.1-flash`, announced for release after
+that date and not served yet, was added at the head of the group at the same price.
+
+## Details
+
+- `deepseek-v4-flash` and `deepseek-v4-flash-vision-exp` in the `deepseek` group moved from CNY
+ 0.1 / 3 / 9 to CNY 0.04 / 2 / 8. DeepSeek publishes the vision revision at V4 Flash's own price,
+ so the two rows stayed equal. At the catalog's 7:1 storage convention that is USD
+ 0.005714 / 0.285714 / 1.142857 per million tokens.
+- The stored number is still the peak one and `offPeakDiscount: DEEPSEEK_OFF_PEAK` is unchanged,
+ so every bucket is still halved outside Beijing weekday 09:00–12:00 and 14:00–18:00, when a
+ price is read rather than when it is written.
+- `deepseek-v4-pro` kept CNY 0.3 / 9 / 27. The adjustment covers the Flash series only.
+- `deepseek-v4.1-flash` (**DeepSeek V4.1 Flash**) was pre-registered at the head of the `deepseek`
+ group: image input per DeepSeek's announcement, the V4 Flash price CNY 0.04 / 2 / 8 on the same
+ off-peak schedule, and a 1,000,000-token context window assumed from `deepseek-v4-flash` until
+ the official model page lists one. Image parts reach it through `@prismshadow/agenthub` 0.4.11,
+ whose DeepSeek client forwards them to every id except the text-only `deepseek-v4-flash` and
+ `deepseek-v4-pro` (see below).
+- The gateway rows reselling DeepSeek — OpenRouter, Fireworks AI, SiliconFlow, TokenDance and the
+ two Qwen groups — were left as they were. Each records what its own seller bills, which is not
+ the vendor's list price.
+- The illustrative `[[models]]` block in the configuration document was moved to the new figures,
+ and the sample list in the models document plus the `unified-llm-api` skill's DeepSeek V4 id
+ list gained the new row.
+
+## The AgentHub dependency moves to 0.4.11
+
+`packages/core` and `packages/cli` move `@prismshadow/agenthub` from `^0.4.10` to `^0.4.11`, the
+release ([agenthub 0.4.11](https://github.com/Prism-Shadow/agenthub/blob/main/changelog/0.4.11/README.md))
+that carries the DeepSeek client's text-only deny-list, the same V4 Flash re-pricing and
+`deepseek-v4.1-flash` in its registry, and the `gpt6` client — so `gpt-6-astra` and
+`openai/gpt-6-astra`, added to this catalog in
+[#654](https://github.com/Prism-Shadow/penguin-harness/pull/654) ahead of that release, become
+routable. The `minimumReleaseAgeExclude` entry in `pnpm-workspace.yaml` moves to the new version
+with it.
+
+## Existing Projects
+
+Presets are copied into `.project_config.toml` when a Project is created and nothing rewrites them
+afterwards, so an existing Project keeps the price it stored; the models page's **sync presets** is
+what brings the new one in. Until it syncs, a Project still holding the old peak price gets neither
+the off-peak split in the cost center nor the `-50%` badge on the models page: both apply only
+while the stored price equals the catalog's current peak price (`tieredRates` in
+`packages/server/src/services/project-config-service.ts`, `discountedPrice` in
+`packages/web/src/features/models/model-grouping.ts`). Because the prices moved and a row was
+added, every existing Project sees the preset-update badge until it syncs or dismisses it.
diff --git a/changelog/unreleased/2026-09-08-deepseek-flash-pricing.zh.md b/changelog/unreleased/2026-09-08-deepseek-flash-pricing.zh.md
new file mode 100644
index 00000000..cfa5576f
--- /dev/null
+++ b/changelog/unreleased/2026-09-08-deepseek-flash-pricing.zh.md
@@ -0,0 +1,49 @@
+# DeepSeek V4 Flash 自 2026-09-10 生效的价格,以及预先登记的 V4.1 Flash 条目
+
+- **Date:** 2026-09-08
+- **Type:** fix
+- **Scope:** `core`, `cli`, `docs`, `skills`
+- **PR:** [#649](https://github.com/Prism-Shadow/penguin-harness/pull/649)
+
+[English](2026-09-08-deepseek-flash-pricing.md)
+
+DeepSeek 调整了 V4 Flash 系列的价格,自北京时间 2026-09-10 12:00 生效。目录中两条 V4 Flash 直连
+条目已于 2026-09-08 重新读取,现记录高峰档 CNY 0.04 / 2 / 8(缓存命中 / 缓存未命中 / 输出,每百万
+Token);其空闲档正好减半——0.02 / 1 / 4——由每条已声明的时段规则在读取时应用。DeepSeek 已宣布
+`deepseek-v4.1-flash` 将在该日期之后发布、目前尚未提供服务,该条目按同一价格加在分组首位。
+
+## 细节
+
+- `deepseek` 分组中的 `deepseek-v4-flash` 与 `deepseek-v4-flash-vision-exp` 由 CNY 0.1 / 3 / 9
+ 改为 CNY 0.04 / 2 / 8。DeepSeek 对视觉版本发布的就是 V4 Flash 自身的价格,因此两条保持一致。
+ 按目录 7:1 的存储约定,即每百万 Token USD 0.005714 / 0.285714 / 1.142857。
+- 落盘的仍是高峰价,`offPeakDiscount: DEEPSEEK_OFF_PEAK` 不变,因此在北京时间周一至周五
+ 9:00–12:00、14:00–18:00 之外各档价格照旧减半——减免发生在读取价格时,而不是写入时。
+- `deepseek-v4-pro` 保持 CNY 0.3 / 9 / 27,本次调整只覆盖 Flash 系列。
+- `deepseek-v4.1-flash`(**DeepSeek V4.1 Flash**)预先登记在 `deepseek` 分组首位:按 DeepSeek 的
+ 发布公告支持图像输入,价格取 V4 Flash 系列的 CNY 0.04 / 2 / 8 并沿用同一空闲时段规则;上下文
+ 窗口在官方模型页给出之前,先按 `deepseek-v4-flash` 的 1,000,000 Token 记。图像部件经
+ `@prismshadow/agenthub` 0.4.11 送达:其 DeepSeek 客户端把图像转发给除 text-only 的
+ `deepseek-v4-flash` 与 `deepseek-v4-pro` 之外的所有 id(见下文)。
+- 转售 DeepSeek 的网关条目——OpenRouter、Fireworks AI、SiliconFlow、TokenDance 以及两个 Qwen
+ 分组——维持原样。每一条记录的是各自卖家的收费,而不是厂商牌价。
+- 配置文档中示意用的 `[[models]]` 代码块同步改为新数字;模型文档的示例清单,以及 `unified-llm-api`
+ 技能中 DeepSeek V4 一行的 id 列表,都加入了新条目。
+
+## AgentHub 依赖升至 0.4.11
+
+`packages/core` 与 `packages/cli` 的 `@prismshadow/agenthub` 从 `^0.4.10` 升至 `^0.4.11`。该版本
+([agenthub 0.4.11](https://github.com/Prism-Shadow/agenthub/blob/main/changelog/0.4.11/README.md))
+带有 DeepSeek 客户端的 text-only 拒绝名单、注册表中同样的 V4 Flash 调价与 `deepseek-v4.1-flash`,
+以及 `gpt6` 客户端——因此 [#654](https://github.com/Prism-Shadow/penguin-harness/pull/654) 中先于
+该版本加入目录的 `gpt-6-astra` 与 `openai/gpt-6-astra` 从此可以路由。`pnpm-workspace.yaml` 中的
+`minimumReleaseAgeExclude` 条目随之移到新版本。
+
+## 存量 Project
+
+预置模型在 Project 创建时复制进 `.project_config.toml`,此后不会被改写,因此存量 Project 保持
+自己存下的价格;新价格只能通过模型页的**同步预置**到达。在同步之前,仍持有旧高峰价的 Project 既
+不会在成本中心得到峰谷拆分,也不会在模型页看到 `-50%` 标记:两者都只在存盘价等于目录当前高峰价时
+生效(`packages/server/src/services/project-config-service.ts` 中的 `tieredRates`、
+`packages/web/src/features/models/model-grouping.ts` 中的 `discountedPrice`)。由于价格发生了
+变化,且新增了一条预置,存量 Project 都会看到预置更新角标,直到它同步或忽略为止。
diff --git a/packages/cli/package.json b/packages/cli/package.json
index 90d86283..acbaa472 100644
--- a/packages/cli/package.json
+++ b/packages/cli/package.json
@@ -36,7 +36,7 @@
"@penguinharness/use-bento-slides": "workspace:*",
"@penguinharness/use-firecrawl": "workspace:*",
"@penguinharness/use-claude-code": "workspace:*",
- "@prismshadow/agenthub": "^0.4.10",
+ "@prismshadow/agenthub": "^0.4.11",
"@prismshadow/penguin-core": "workspace:*",
"@prismshadow/penguin-server": "workspace:*",
"commander": "^13.0.0",
diff --git a/packages/core/package.json b/packages/core/package.json
index 1253b5d7..3983a491 100644
--- a/packages/core/package.json
+++ b/packages/core/package.json
@@ -65,7 +65,7 @@
"@penguinharness/use-bento-slides": "workspace:*",
"@penguinharness/use-firecrawl": "workspace:*",
"@penguinharness/use-claude-code": "workspace:*",
- "@prismshadow/agenthub": "^0.4.10",
+ "@prismshadow/agenthub": "^0.4.11",
"arktype": "^2.2.3",
"smol-toml": "^1.3.0",
"yaml": "^2.5.0"
diff --git a/packages/core/src/state/model-catalog.ts b/packages/core/src/state/model-catalog.ts
index ac362b28..c2edb7a2 100644
--- a/packages/core/src/state/model-catalog.ts
+++ b/packages/core/src/state/model-catalog.ts
@@ -7,8 +7,11 @@
* 2026-08-21; the TokenDance group: 2026-08-25, its glm-5.3-flash row: 2026-08-26, its
* qwen3.8-flash row: 2026-08-27 and its running promotions plus the hy4-preview rows
* (TokenDance + OpenRouter): 2026-08-28; the GLM-5.3 Flash rows (direct + OpenRouter) and
- * the direct qwen3.8-flash: 2026-08-26; the vLLM group: 2026-09-03; the GPT-6 Astra rows
- * (direct + OpenRouter): 2026-09-09 — per each provider's docs).
+ * the direct qwen3.8-flash: 2026-08-26; the vLLM group: 2026-09-03; the direct DeepSeek V4
+ * Flash rows (flash and flash-vision-exp): 2026-09-08, for the official adjustment effective
+ * 2026-09-10; the GPT-6 Astra rows (direct + OpenRouter): 2026-09-09; the pre-registered
+ * deepseek-v4.1-flash row: 2026-09-09, from DeepSeek's release announcement — per each
+ * provider's docs).
* Docs: packages/docs/content/models.{zh,en}.md (site path /docs/models) documents the
* provider groups and credential resolution described here.
*
@@ -462,18 +465,39 @@ export function effectivePricing(
* takes the group sequence from MODEL_PROVIDERS.
*/
export const MODEL_CATALOG: ModelCatalogEntry[] = [
- // -- DeepSeek (official CNY pricing: cache hit / cache miss / output). Re-read 2026-08-18
- // from api-docs.deepseek.com/quick_start/pricing after the official price increase
- // introduced time-based tiers. The rows store the PEAK tier and declare
- // DEEPSEEK_OFF_PEAK, which halves every bucket outside Beijing weekday 09:00-12:00 and
- // 14:00-18:00 — so both tiers are billed at the rate actually in force, rather than one of
- // them being approximated by the other. --
+ // -- DeepSeek (official CNY pricing: cache hit / cache miss / output). Prices re-read
+ // 2026-09-08 from api-docs.deepseek.com/quick_start/pricing, covering the official price
+ // adjustment effective 2026-09-10 12:00 Beijing: the V4 Flash rows (flash and
+ // flash-vision-exp) are on the peak tier CNY 0.04 / 2 / 8, exactly double the off-peak
+ // 0.02 / 1 / 4; deepseek-v4-pro sits outside that adjustment, at 0.3 / 9 / 27. The rows
+ // store the PEAK tier and declare DEEPSEEK_OFF_PEAK, which halves every bucket outside
+ // Beijing weekday 09:00-12:00 and 14:00-18:00 — so both tiers are billed at the rate
+ // actually in force, rather than one of them being approximated by the other.
+ // deepseek-v4.1-flash leads the group ahead of its launch: DeepSeek has announced it and
+ // does not serve it yet, and it carries the V4 Flash series price and schedule. --
+ {
+ // Announced for release after 2026-09-10 and not served yet as of 2026-09-09; the row is
+ // registered ahead of the launch. The announcement gives it image input by default, and
+ // the V4 Flash series price: the peak tier is stored and the schedule declared, exactly
+ // like its siblings. The context window is assumed equal to deepseek-v4-flash until the
+ // official model page lists one. Image parts reach it through @prismshadow/agenthub
+ // ^0.4.11, whose DeepSeek client forwards them to every id except the text-only
+ // deepseek-v4-flash and deepseek-v4-pro (0.4.10 forwarded them to ids containing
+ // "vision" alone).
+ modelId: "deepseek-v4.1-flash",
+ displayName: "DeepSeek V4.1 Flash",
+ provider: "deepseek",
+ contextWindow: 1000000,
+ pricing: cny(0.04, 2, 8),
+ offPeakDiscount: DEEPSEEK_OFF_PEAK,
+ supportsVision: true,
+ },
{
modelId: "deepseek-v4-flash",
displayName: "DeepSeek V4 Flash",
provider: "deepseek",
contextWindow: 1000000,
- pricing: cny(0.1, 3, 9),
+ pricing: cny(0.04, 2, 8),
offPeakDiscount: DEEPSEEK_OFF_PEAK,
supportsVision: false,
},
@@ -484,7 +508,7 @@ export const MODEL_CATALOG: ModelCatalogEntry[] = [
displayName: "DeepSeek V4 Flash Vision Exp",
provider: "deepseek",
contextWindow: 1000000,
- pricing: cny(0.1, 3, 9),
+ pricing: cny(0.04, 2, 8),
offPeakDiscount: DEEPSEEK_OFF_PEAK,
supportsVision: true,
},
diff --git a/packages/core/test/model-catalog.test.ts b/packages/core/test/model-catalog.test.ts
index 5d0b1e3c..2a065a7b 100644
--- a/packages/core/test/model-catalog.test.ts
+++ b/packages/core/test/model-catalog.test.ts
@@ -242,11 +242,14 @@ describe("model-catalog", () => {
expect(catalogEntryFor("qwen-token-plan", "glm-5.2")?.contextWindow).toBe(1048576);
expect(catalogEntryFor("deepseek", "deepseek-v4-pro")?.provider).toBe("deepseek");
// The vision revision is a model of its own in both the direct group and on OpenRouter,
- // and it is the only vision-capable DeepSeek row in either.
+ // and on OpenRouter it is the only vision-capable DeepSeek row. In the direct group the
+ // pre-registered deepseek-v4.1-flash reads images too, so the flag is per row rather than
+ // a property of the id's spelling.
expect(catalogEntryFor("deepseek", "deepseek-v4-flash-vision-exp")?.supportsVision).toBe(true);
expect(
catalogEntryFor("openrouter", "deepseek/deepseek-v4-flash-vision-exp")?.supportsVision,
).toBe(true);
+ expect(catalogEntryFor("deepseek", "deepseek-v4.1-flash")?.supportsVision).toBe(true);
expect(catalogEntryFor("deepseek", "deepseek-v4-flash")?.supportsVision).toBe(false);
expect(catalogEntryFor("qwen-token-plan", "deepseek-v4-pro")?.provider).toBe("qwen-token-plan");
expect(catalogEntryFor("minimax", "MiniMax-M3")?.displayName).toBe("MiniMax M3");
@@ -782,12 +785,13 @@ describe("model-catalog", () => {
it("DeepSeek and Kimi are initialized from official CNY prices (stored in USD; x7 recovers the official price)", () => {
const cnyOf = (usdV: number) => Math.round(usdV * 7 * 1000) / 1000;
- // DeepSeek rows carry the official PEAK tier — re-read 2026-08-18 after the official price
- // increase introduced time-based tiers. The off-peak tier is exactly half, and is applied
- // from the row's schedule rather than stored (see the off-peak schedules block below).
+ // DeepSeek rows carry the official PEAK tier; the off-peak tier is exactly half, and is
+ // applied from the row's schedule rather than stored (see the off-peak schedules block
+ // below). The flash rows were re-read 2026-09-08 for the official price adjustment
+ // effective 2026-09-10; V4 Pro sits outside that adjustment.
const flash = catalogEntryFor("deepseek", "deepseek-v4-flash")!.pricing!;
expect([cnyOf(flash.cache_read), cnyOf(flash.cache_write), cnyOf(flash.output)]).toEqual([
- 0.1, 3, 9,
+ 0.04, 2, 8,
]);
const pro = catalogEntryFor("deepseek", "deepseek-v4-pro")!.pricing!;
expect([cnyOf(pro.cache_read), cnyOf(pro.cache_write), cnyOf(pro.output)]).toEqual([
@@ -916,6 +920,9 @@ describe("model-catalog", () => {
describe("resolveModelEnv (PRN-021: env fallback resolved by AgentHub routing rules)", () => {
it("first-party model ids route to the provider client's env var", () => {
expect(resolveModelEnv("deepseek-v4-pro")?.envKey).toBe("DEEPSEEK_API_KEY");
+ // The dotted V4.1 spelling still carries the deepseek-v4 substring AutoLLMClient routes on.
+ expect(resolveModelEnv("deepseek-v4.1-flash")?.envKey).toBe("DEEPSEEK_API_KEY");
+ expect(resolveModelEnv("deepseek-v4.1-flash")?.envBaseUrlKey).toBe("DEEPSEEK_BASE_URL");
expect(resolveModelEnv("claude-opus-4-8")?.envKey).toBe("ANTHROPIC_API_KEY");
expect(resolveModelEnv("claude-sonnet-4-6")?.envKey).toBe("ANTHROPIC_API_KEY");
expect(resolveModelEnv("gemini-3.5-flash")?.envKey).toBe("GEMINI_API_KEY");
@@ -1210,11 +1217,11 @@ describe("off-peak schedules", () => {
5,
);
}
- // The published peak figures themselves: CNY 0.1 / 3 / 9 per million, at the catalog's 7:1
- // display convention, which is DeepSeek's off-peak 0.05 / 1.5 / 4.5 doubled.
+ // The published peak figures themselves: CNY 0.04 / 2 / 8 per million, at the catalog's 7:1
+ // display convention, which is DeepSeek's off-peak 0.02 / 1 / 4 doubled.
const flash = MODEL_CATALOG.find((m) => m.modelId === "deepseek-v4-flash")!.pricing!;
expect([flash.cache_read, flash.cache_write, flash.output]).toEqual([
- 0.014286, 0.428571, 1.285714,
+ 0.005714, 0.285714, 1.142857,
]);
});
diff --git a/packages/docs/content/configuration.en.md b/packages/docs/content/configuration.en.md
index 639f3ac1..aabb2a4b 100644
--- a/packages/docs/content/configuration.en.md
+++ b/packages/docs/content/configuration.en.md
@@ -84,9 +84,9 @@ api_key = "sk-..."
[models.pricing]
unit = "usd_per_mtok"
-cache_read = 0.003571
-cache_write = 0.428571
-output = 0.857143
+cache_read = 0.005714
+cache_write = 0.285714
+output = 1.142857
```
`pricing.unit` is currently always `usd_per_mtok` (USD per million tokens); the three buckets map onto `token_usage`'s three counters.
diff --git a/packages/docs/content/configuration.zh.md b/packages/docs/content/configuration.zh.md
index 4725e184..eb4810f0 100644
--- a/packages/docs/content/configuration.zh.md
+++ b/packages/docs/content/configuration.zh.md
@@ -84,9 +84,9 @@ api_key = "sk-..."
[models.pricing]
unit = "usd_per_mtok"
-cache_read = 0.003571
-cache_write = 0.428571
-output = 0.857143
+cache_read = 0.005714
+cache_write = 0.285714
+output = 1.142857
```
`pricing.unit` 目前固定为 `usd_per_mtok`(USD 每百万 Token);三档对应 `token_usage` 的三个计数桶。
diff --git a/packages/docs/content/models.en.md b/packages/docs/content/models.en.md
index 2c85c1c8..251a3ab6 100644
--- a/packages/docs/content/models.en.md
+++ b/packages/docs/content/models.en.md
@@ -78,7 +78,7 @@ The gateway groups (openrouter / fireworks / siliconflow / tokendance / qwen-pay
The preset catalog also carries OpenRouter's free tier: the `:free` model variant `nvidia/nemotron-3-ultra-550b-a55b:free` and the `openrouter/free` unified Free Models Router. They cost nothing, but are subject to OpenRouter's free-tier rate limits and data policy.
-Some models in the preset catalog: deepseek-v4-pro / deepseek-v4-flash / deepseek-v4-flash-vision-exp (the DeepSeek group's only vision-capable model), MiniMax-M3, gemini-3.7-flash, claude-opus-5 / claude-opus-4-8 / claude-sonnet-5, gpt-6-astra / gpt-5.6 / gpt-5.5, glm-5.3 / glm-5.3-flash, kimi-k3, qwen3.8-max / qwen3.8-flash (not exhaustive). The whole OpenAI line-up is listed twice — directly (your own OpenAI key, list prices) and on OpenRouter as `openai/<id>` (the gateway's rates, which follow its running promotions). DeepSeek's direct-group rows record the official peak tier and declare its off-peak schedule: outside Beijing weekday 9:00–12:00 and 14:00–18:00 every bucket is halved, which the models page marks with a `-50%` tag and the cost center bills at. The stored price is always the peak one, so what is on disk does not depend on the hour a Project was created or re-synced in. `glm-5.3-flash` appears three times, and all three rows accept images: AgentHub's GLM client forwards image parts for this one GLM id (every other GLM id refuses them), while the OpenRouter row `z-ai/glm-5.3-flash` and the TokenDance row go through the generic OpenAI-compatible client, which carries them for any id. What the three rows do not share is the price: each records what its own seller charges, so they disagree while a promotion is running.
+Some models in the preset catalog: deepseek-v4.1-flash / deepseek-v4-pro / deepseek-v4-flash / deepseek-v4-flash-vision-exp (deepseek-v4.1-flash and deepseek-v4-flash-vision-exp are the DeepSeek group's vision-capable models), MiniMax-M3, gemini-3.7-flash, claude-opus-5 / claude-opus-4-8 / claude-sonnet-5, gpt-6-astra / gpt-5.6 / gpt-5.5, glm-5.3 / glm-5.3-flash, kimi-k3, qwen3.8-max / qwen3.8-flash (not exhaustive). The whole OpenAI line-up is listed twice — directly (your own OpenAI key, list prices) and on OpenRouter as `openai/<id>` (the gateway's rates, which follow its running promotions). DeepSeek's direct-group rows record the official peak tier and declare its off-peak schedule: outside Beijing weekday 9:00–12:00 and 14:00–18:00 every bucket is halved, which the models page marks with a `-50%` tag and the cost center bills at. The stored price is always the peak one, so what is on disk does not depend on the hour a Project was created or re-synced in. `glm-5.3-flash` appears three times, and all three rows accept images: AgentHub's GLM client forwards image parts for this one GLM id (every other GLM id refuses them), while the OpenRouter row `z-ai/glm-5.3-flash` and the TokenDance row go through the generic OpenAI-compatible client, which carries them for any id. What the three rows do not share is the price: each records what its own seller charges, so they disagree while a promotion is running.
TokenDance rows record the gateway's list price and, where a promotion is running, the rate off it. Six models are discounted today — `deepseek-v4-flash-0731`, `deepseek-v4-pro-0813` and `glm-5.3-flash` at 50% off, `kimi-k3` at 20%, `glm-5.3` and `qwen3.8-max` at 10%. Their model cards show the price being billed right now, with the rate as a badge, and a Project is preset with the **discounted** price, so the cost center charges what the gateway charges. Prices you edit yourself keep the discount decoration off the card: the figure is then yours, not the gateway's.
diff --git a/packages/docs/content/models.zh.md b/packages/docs/content/models.zh.md
index 09d9fa65..6968a179 100644
--- a/packages/docs/content/models.zh.md
+++ b/packages/docs/content/models.zh.md
@@ -78,7 +78,7 @@ api_key = "sk-..."
预置目录还收录了 OpenRouter 的免费档:`:free` 模型变体 `nvidia/nemotron-3-ultra-550b-a55b:free` 与统一路由 `openrouter/free`(Free Models Router),零成本可用,但受 OpenRouter 免费档速率限制与数据政策约束。
-预置目录中的部分模型:deepseek-v4-pro / deepseek-v4-flash / deepseek-v4-flash-vision-exp(DeepSeek 分组中唯一支持图像输入的模型)、MiniMax-M3、gemini-3.7-flash、claude-opus-5 / claude-opus-4-8 / claude-sonnet-5、gpt-6-astra / gpt-5.6 / gpt-5.5、glm-5.3 / glm-5.3-flash、kimi-k3、qwen3.8-max / qwen3.8-flash 等(非完整清单)。OpenAI 全系列都收录了两份——直连(用自己的 OpenAI Key,记牌价)与 OpenRouter 上的 `openai/<id>`(记网关实际计费价,会随其促销浮动)。DeepSeek 直连分组的条目记录官方高峰档,并声明其空闲时段规则:在北京时间周一至周五 9:00–12:00、14:00–18:00 之外,各档价格减半——模型页以 `-50%` 标记,成本中心按此计价。存盘的始终是高峰价,因此磁盘上的数字不会随 Project 创建或同步预置的时刻而变。`glm-5.3-flash` 收录了三份,三条都支持图像输入:AgentHub 的 GLM 客户端只为这一个 GLM id 转发图像部件(其余 GLM id 一律拒绝),而 OpenRouter 上的 `z-ai/glm-5.3-flash` 与 TokenDance 分组的同名条目走通用 OpenAI 兼容客户端,对任何 id 都能携带图片。三条不一致的是价格:每一条都记录各自卖家的收费,因此促销期间三者不同。
+预置目录中的部分模型:deepseek-v4.1-flash / deepseek-v4-pro / deepseek-v4-flash / deepseek-v4-flash-vision-exp(其中 deepseek-v4.1-flash 与 deepseek-v4-flash-vision-exp 是 DeepSeek 分组中支持图像输入的两条)、MiniMax-M3、gemini-3.7-flash、claude-opus-5 / claude-opus-4-8 / claude-sonnet-5、gpt-6-astra / gpt-5.6 / gpt-5.5、glm-5.3 / glm-5.3-flash、kimi-k3、qwen3.8-max / qwen3.8-flash 等(非完整清单)。OpenAI 全系列都收录了两份——直连(用自己的 OpenAI Key,记牌价)与 OpenRouter 上的 `openai/<id>`(记网关实际计费价,会随其促销浮动)。DeepSeek 直连分组的条目记录官方高峰档,并声明其空闲时段规则:在北京时间周一至周五 9:00–12:00、14:00–18:00 之外,各档价格减半——模型页以 `-50%` 标记,成本中心按此计价。存盘的始终是高峰价,因此磁盘上的数字不会随 Project 创建或同步预置的时刻而变。`glm-5.3-flash` 收录了三份,三条都支持图像输入:AgentHub 的 GLM 客户端只为这一个 GLM id 转发图像部件(其余 GLM id 一律拒绝),而 OpenRouter 上的 `z-ai/glm-5.3-flash` 与 TokenDance 分组的同名条目走通用 OpenAI 兼容客户端,对任何 id 都能携带图片。三条不一致的是价格:每一条都记录各自卖家的收费,因此促销期间三者不同。
TokenDance 分组的条目记录该网关的牌价,并在有促销时记录折扣率。当前有六个模型处于折扣中——`deepseek-v4-flash-0731`、`deepseek-v4-pro-0813` 与 `glm-5.3-flash` 五折,`kimi-k3` 八折,`glm-5.3` 与 `qwen3.8-max` 九折。它们的模型卡片显示当前实际计费的那个价格,并以徽标标出折扣率;新建 Project 预置的是**折后价**,因此成本中心按网关实际收费计价。自行修改过价格的条目不再显示折扣标记:此时那个数字属于你,而不是网关。
diff --git a/plugins/agent-development/skills/unified-llm-api/SKILL.md b/plugins/agent-development/skills/unified-llm-api/SKILL.md
index 2fd3b649..e628d362 100644
--- a/plugins/agent-development/skills/unified-llm-api/SKILL.md
+++ b/plugins/agent-development/skills/unified-llm-api/SKILL.md
@@ -61,7 +61,7 @@ Use exact model ids. If an id is not in the table below and the user has not giv
| Kimi K3 | `kimi-k3` | OpenRouter `moonshotai/kimi-k3` |
| Kimi K2.7 Code | — | SiliconFlow `moonshotai/Kimi-K2.7-Code`; Fireworks AI `accounts/fireworks/models/kimi-k2p7-code` |
| Kimi K2.6 | `kimi-k2.6` | OpenRouter `moonshotai/kimi-k2.6`; SiliconFlow `Pro/moonshotai/Kimi-K2.6` |
-| DeepSeek V4 | `deepseek-v4-pro`, `deepseek-v4-flash`, `deepseek-v4-flash-vision-exp` | OpenRouter `deepseek/deepseek-v4-pro-0813`, `deepseek/deepseek-v4-pro`, `deepseek/deepseek-v4-flash`, `deepseek/deepseek-v4-flash-0731`, `deepseek/deepseek-v4-flash-vision-exp`; Fireworks AI `accounts/fireworks/models/deepseek-v4-flash-0731`; SiliconFlow `deepseek-ai/DeepSeek-V4-Pro`, `deepseek-ai/DeepSeek-V4-Flash` |
+| DeepSeek V4 | `deepseek-v4.1-flash`, `deepseek-v4-pro`, `deepseek-v4-flash`, `deepseek-v4-flash-vision-exp` | OpenRouter `deepseek/deepseek-v4-pro-0813`, `deepseek/deepseek-v4-pro`, `deepseek/deepseek-v4-flash`, `deepseek/deepseek-v4-flash-0731`, `deepseek/deepseek-v4-flash-vision-exp`; Fireworks AI `accounts/fireworks/models/deepseek-v4-flash-0731`; SiliconFlow `deepseek-ai/DeepSeek-V4-Pro`, `deepseek-ai/DeepSeek-V4-Flash` |
| GLM 5.3 | `glm-5.3`, `glm-5.3-flash` | OpenRouter `z-ai/glm-5.3`, `z-ai/glm-5.3-flash` |
| GLM 5.2 | `glm-5.2` | OpenRouter `z-ai/glm-5.2`; SiliconFlow `zai-org/GLM-5.2` |
| GLM 5.1 | `glm-5.1` | — |
diff --git a/pnpm-lock.yaml b/pnpm-lock.yaml
index 691bd88b..57394e77 100644
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
@@ -88,8 +88,8 @@ importers:
specifier: workspace:*
version: link:../../plugins/use-firecrawl
'@prismshadow/agenthub':
- specifier: ^0.4.10
+ specifier: ^0.4.11
'@prismshadow/penguin-core':
specifier: workspace:*
version: link:../core
@@ -167,8 +167,8 @@ importers:
specifier: workspace:*
version: link:../../plugins/use-firecrawl
'@prismshadow/agenthub':
- specifier: ^0.4.10
+ specifier: ^0.4.11
arktype:
specifier: ^2.2.3
version: 2.2.3
@@ -1254,8 +1254,8 @@ packages:
engines: {node: '>=18'}
hasBin: true
- '@prismshadow/[email protected]':
- resolution: {integrity: sha512-zrBF/AVCJlLfjTFYRcxl9awksBNVSzkKqseTHZfEDIifg1ng5+a/LtWnABf8u3WcWlntU14JwdJiy1/4tgMqLw==}
+ '@prismshadow/[email protected]':
+ resolution: {integrity: sha512-BuLjkGhlihbtDQjyFkWhrBlTAxqi9xNqfhNeP0BEQy8it2D/DxoLqhs74EhmMUlWRRoxNEIJzlQFnLWGWzUUwg==}
'@prismshadow/penguin-cli@file:packages/cli':
resolution: {directory: packages/cli, type: directory}
@@ -5001,7 +5001,7 @@ snapshots:
dependencies:
playwright: 1.61.1
dependencies:
'@anthropic-ai/bedrock-sdk': 0.26.4([email protected])
'@anthropic-ai/sdk': 0.91.1([email protected])
@@ -5016,7 +5016,7 @@ snapshots:
- ws
- zod
dependencies:
'@anthropic-ai/bedrock-sdk': 0.26.4([email protected])
'@anthropic-ai/sdk': 0.91.1([email protected])
@@ -5046,7 +5046,7 @@ snapshots:
'@penguinharness/use-bento-slides': file:plugins/use-bento-slides
'@penguinharness/use-claude-code': file:plugins/use-claude-code
'@penguinharness/use-firecrawl': file:plugins/use-firecrawl
- '@prismshadow/agenthub': 0.4.10([email protected])([email protected])
+ '@prismshadow/agenthub': 0.4.11([email protected])([email protected])
'@prismshadow/penguin-core': file:packages/core([email protected])([email protected])
'@prismshadow/penguin-server': file:packages/server([email protected])
commander: 13.1.0
@@ -5077,7 +5077,7 @@ snapshots:
'@penguinharness/use-bento-slides': file:plugins/use-bento-slides
'@penguinharness/use-claude-code': file:plugins/use-claude-code
'@penguinharness/use-firecrawl': file:plugins/use-firecrawl
- '@prismshadow/agenthub': 0.4.10([email protected])([email protected])
+ '@prismshadow/agenthub': 0.4.11([email protected])([email protected])
arktype: 2.2.3
smol-toml: 1.6.1
yaml: 2.9.0
@@ -5104,7 +5104,7 @@ snapshots:
'@penguinharness/use-bento-slides': file:plugins/use-bento-slides
'@penguinharness/use-claude-code': file:plugins/use-claude-code
'@penguinharness/use-firecrawl': file:plugins/use-firecrawl
- '@prismshadow/agenthub': 0.4.10([email protected])([email protected])([email protected])
+ '@prismshadow/agenthub': 0.4.11([email protected])([email protected])([email protected])
arktype: 2.2.3
smol-toml: 1.6.1
yaml: 2.9.0
diff --git a/pnpm-workspace.yaml b/pnpm-workspace.yaml
index 221bef67..ef7a2db8 100644
--- a/pnpm-workspace.yaml
+++ b/pnpm-workspace.yaml
@@ -37,4 +37,4 @@ allowBuilds:
# Every bump lands inside the age gate, so the entry names the pinned version and moves
# with each bump rather than being removed.
minimumReleaseAgeExclude:
- - "@prismshadow/[email protected]"
+ - "@prismshadow/[email protected]"