BofAI · Will-Guan · Aug 14, 2026 · Aug 14, 2026
diff --git a/docs/llmservice/models/deepseek-v4-flash.md b/docs/llmservice/models/deepseek-v4-flash.md
@@ -1,9 +1,18 @@
-# DeepSeek V4 Flash
+# DeepSeek-V4-Flash
 
 ## Overview
 
 DeepSeek-V4-Flash is DeepSeek's high-efficiency open-source language model, released alongside V4-Pro on April 24, 2026 under the MIT License. With 284 billion total parameters and only 13 billion active parameters, it delivers performance within striking distance of V4-Pro at roughly 3.1x lower cost, making it one of the most cost-effective models available.
 
+:::tip 🔥 Limited-Time Free Access
+Offer starts August 17, 2026, and applies to DeepSeek-V4-Flash on B.AI Chat and API.
+
+* **Chat:** Usage is billed at `0 Credits` during the offer.
+* **API:** Usage is billed at `0 Credits` during the offer. No per-request, input, cache write, cache read, or output token fees apply.
+
+After the offer ends, the model will return to standard pricing. Offer end time, eligibility, actual charges, and final billing are subject to the platform display.
+:::
+
 ## Key Features
 
 * **Ultra-Efficient Architecture**: 284B total parameters with just 13B activated per forward pass, resulting in a compact 160GB download that runs on significantly less hardware than frontier models while maintaining strong performance.
@@ -40,7 +49,7 @@ DeepSeek-V4-Flash is DeepSeek's high-efficiency open-source language model, rele
 
 | Model | Input (Credits/Token) | Cache Write (Credits/Token) | Cache Read (Credits/Token) | Output (Credits/Token) | Web Search (Credits/Use) | Billing Notes |
 | :--- | --------------------: | --------------------------: | -------------------------: | ---------------------: | -----------------------: | :--- |
-| **DeepSeek V4 Flash** | `0.28` | `0.28` | `0.0056` | `0.56` | `-` | - |
+| **DeepSeek-V4-Flash** | `0.28` | `0.28` | `0.0056` | `0.56` | `-` | - |
 
 :::info Pricing note
 Prices shown in the documentation are B.AI standard reference prices for base billing purposes. B.AI may provide lower actual usage costs through top-up bonuses and account benefits. Specific prices, bonus Credits, and account benefits are subject to the platform display and final billing records.

diff --git a/docs/llmservice/models/glm-5-2.md b/docs/llmservice/models/glm-5-2.md
@@ -4,7 +4,7 @@
 
 GLM-5.2 is a GLM-family text foundation model developed by Z.AI and released on June 16, 2026. It is positioned for long-horizon coding and engineering tasks, with a 1M-token context window, 128K maximum output, and a `reasoning_effort` control for adjusting reasoning depth.
 
-:::tip Limited-time offer: GLM-5.2 at 40% off
+:::note Limited-time offer: GLM-5.2 at 40% off
 Offer starts August 12, 2026.
 
 **Eligibility:** This offer applies to GLM-5.2 requests made through the B.AI API and B.AI web app.

diff --git a/docs/llmservice/models/glm-5-3.md b/docs/llmservice/models/glm-5-3.md
@@ -0,0 +1,59 @@
+# GLM-5.3
+
+## Overview
+
+GLM-5.3 is a text foundation model released by Z.AI on August 14, 2026 for complex coding, long-horizon agent tasks, and cybersecurity work. It uses the same base model as GLM-5.2; Z.AI attributes the update's gains to scaled post-training across more environments, more diverse tasks, and longer trajectories.
+
+:::note 🎁 Limited-time offer: GLM-5.3 at 10% off
+Offer starts August 14, 2026.
+
+**Eligibility:** This offer applies to GLM-5.3 requests made through the B.AI API and B.AI web app.
+
+For a limited time, eligible requests are billed at 90% of the standard reference price: Input `1.26`, Cache Write `1.26`, Cache Read `0.234`, and Output `3.96` Credits/Token.
+
+This page continues to show standard reference prices. Offer end time, eligibility, actual settlement prices, and final billing are subject to the platform display.
+:::
+
+## Key Features
+
+* **Scaled Post-Training**: Retains the GLM-5.2 base model while expanding reinforcement-learning environments, task diversity, and training compute for long-horizon professional workflows.
+* **Configurable Always-On Reasoning**: Supports `low`, `high`, and `max` reasoning effort, with `max` as the default. Native GLM-5.3 requests require thinking to remain enabled.
+* **Coding and Agent Performance**: Z.AI reports 28.3 on Terminal-Bench 3.0, 66.9 on DeepSWE v1.1, 78.1 on FrontierSWE, and 48.2 on AutomationBench v1.0.6.
+* **Cybersecurity Evaluation**: Z.AI reports 84.5 on CyberGym, 54.4 on ExploitBench, and 105/130 completed ExploitGym tasks under normalized two-hour/six-hour budgets.
+* **1M-Token Coding Workflows**: The GLM Coding Plan supports a 1M-token context. Claude Code users enable it with the `glm-5.3[1m]` model name and a 1,000,000-token auto-compaction window.
+
+## Best Use Cases
+
+* **Complex Software Engineering**: Repository-scale implementation, debugging, performance optimization, and test-fix-verify loops that require sustained work across many files and tools.
+* **Long-Horizon Coding Agents**: ZCode, Claude Code, Codex, OpenCode, Cline, and other configurable agents that can use Z.AI's Anthropic-compatible or OpenAI-compatible Coding Plan endpoints.
+* **Authorized Security Research**: Vulnerability discovery, validation, and exploit-chain analysis in systems the operator is permitted to test.
+* **Tool-Driven Professional Workflows**: Multi-step engineering and research tasks with executable environments, objective verification, and iterative feedback.
+
+## Capabilities and Limitations
+
+| Capability | Description |
+| :--- | :--- |
+| **Reasoning** | Thinking is always enabled. `reasoning_effort` supports `low`, `high`, and `max`; the default is `max`, which Z.AI recommends for coding tasks. |
+| **Creative Writing** | General text generation is available. |
+| **Coding** | Z.AI reports Terminal-Bench 2.1: 88.2, Terminal-Bench 3.0: 28.3, DeepSWE v1.1: 66.9, NL2Repo: 58.0, FrontierSWE: 78.1, and SWE-Marathon v1.1: 42.5. |
+| **Multimodal** | Text input and text output. |
+| **Response Speed** | Not published as a guaranteed API rate. |
+| **Context Window** | Up to 1M tokens in GLM Coding Plan workflows. Claude Code requires the `glm-5.3[1m]` suffix to enable the 1M-token mode. |
+| **Max Output** | Not published as a general API limit. |
+| **Tool Use** | Designed and evaluated in tool-using coding-agent harnesses; available to Coding Plan users through ZCode and configurable Anthropic-compatible or OpenAI-compatible agents. |
+| **Multilingual** | Natural-language prompting is supported. |
+
+### Known Limitations
+
+* Native GLM-5.3 API requests do not support `thinking.type: "disabled"`; applications must enable thinking and can choose `reasoning_effort: "low"` for lighter reasoning. The Coding Plan compatibility layer may automatically map disabled thinking to `low`.
+* The model is text-only; image, audio, and video inputs are not advertised for GLM-5.3.
+
+## Credits Usage
+
+| Model | Input (Credits/Token) | Cache Write (Credits/Token) | Cache Read (Credits/Token) | Output (Credits/Token) | Web Search (Credits/Use) | Billing Notes |
+| :--- | --------------------: | --------------------------: | -------------------------: | ---------------------: | -----------------------: | :--- |
+| **GLM-5.3** | `1.40` | `1.40` | `0.26` | `4.40` | `-` | - |
+
+:::info Pricing note
+Prices shown in the documentation are B.AI standard reference prices for base billing purposes. B.AI may provide lower actual usage costs through top-up bonuses and account benefits. Specific prices, bonus Credits, and account benefits are subject to the platform display and final billing records.
+:::
diff --git a/docs/llmservice/models/grok-4.6.md b/docs/llmservice/models/grok-4.6.md
@@ -0,0 +1,53 @@
+# Grok 4.6
+
+## Overview
+
+Grok 4.6 is a SpaceXAI reasoning model in the Grok 4 family, released on August 12, 2026 for coding, long-running agents, interactive and visual projects, and knowledge work. Available through the API as `grok-4.6`, it supports text and image input, text output, a 500,000-token context window, and four configurable reasoning-effort levels.
+
+## Key Features
+
+* **Long-Horizon Agentic Work**: Trained on agentic reinforcement-learning tasks across knowledge work, general coding, kernel optimization, web development, and computer-aided design, with an emphasis on sustaining complex work across many steps.
+* **Configurable Reasoning**: Supports `low`, `medium`, `high`, and `xhigh` reasoning effort, with `high` as the default. Reasoning cannot be disabled, and the API can stream summarized reasoning content.
+* **Coding and Knowledge-Work Evaluations**: SpaceXAI reports 69.9% on CursorBench v3.2, 65.9% on DeepSWE v1.1, 61.3% on FrontierCode v1.1 Extended, and 57.5% on APEX-Agents.
+* **Long Multimodal Context**: Accepts text and image input within a 500,000-token context window and returns text output. Prompts at or above 200,000 tokens use long-context pricing.
+* **Structured Agent Workflows**: Supports function calling, parallel tool calls, structured outputs, and built-in tools such as web search and X search.
+
+## Best Use Cases
+
+* **Repository-Scale Software Engineering**: Implementing features, debugging, refactoring, and validating changes across large codebases or long-running coding sessions.
+* **Interactive Product Prototypes**: Turning a broad product idea into a working application, including its structure, visual language, and core interactions, then iterating from feedback.
+* **Tool-Using Research Agents**: Combining reasoning with web or X search, custom functions, structured outputs, and repeated verification steps.
+* **Technical Knowledge Work**: Analyzing documents and images or producing technical artifacts across science, engineering, mathematics, and other professional domains.
+
+## Capabilities and Limitations
+
+| Capability | Description |
+| :--- | :--- |
+| **Reasoning** | Supports `low`, `medium`, `high`, and `xhigh` reasoning effort; `high` is the default and reasoning cannot be disabled. Official results include 61 on the AA Intelligence Index and 1,753 on GDPVal-AA v2. |
+| **Coding** | Designed for long-running agentic coding and technical workflows. Official results include 69.9% on CursorBench v3.2, 65.9% on DeepSWE v1.1, 61.3% on FrontierCode v1.1 Extended, and 26% on Terminal-Bench v3.0. |
+| **Creative Writing** | Supports general text generation and document creation. |
+| **Multimodal** | Text and image input with text output. |
+| **Response Speed** | SpaceXAI has not published a model-specific latency or tokens-per-second figure. |
+| **Context Window** | 500,000 tokens. Prompts at or above 200,000 tokens are billed at the long-context rates for all tokens in the request. |
+| **Max Output** | SpaceXAI's release notes state that the model has no text output limit. |
+| **Tool Use** | Function calling, parallel tool calls, structured outputs, and built-in tools including web search and X search. The Responses API also supports code execution, file or collections search, and Remote MCP tools. |
+| **Multilingual** | Natural-language prompting is supported. |
+| **Knowledge Cutoff** | February 1, 2026. |
+
+### Known Limitations
+
+* `logprobs` and `top_logprobs` are not supported by Grok 4.20 and newer models. Reasoning requests also reject `presencePenalty`, `frequencyPenalty`, and `stop`.
+* SpaceXAI has not published a model-specific throughput figure, multilingual evaluation, or separate numeric output-token ceiling.
+
+## Credits Usage
+
+| Model and Context Tier | Input (Credits/Token) | Cache Write (Credits/Token) | Cache Read (Credits/Token) | Output (Credits/Token) |
+| :--- | --------------------: | --------------------------: | -------------------------: | ---------------------: |
+| **Grok 4.6** | `2.00` | `2.00` | `0.50` | `6.00` |
+| **Grok 4.6** (>=200K prompt tokens) | `4.00` | `2.00` | `1.00` | `12.00` |
+
+* Once a prompt reaches 200,000 tokens, long-context rates apply to all input, cached input, reasoning, and output tokens in the request. Reasoning tokens are billed at the output-token rate.
+
+:::info Pricing note
+Prices shown in the documentation are B.AI standard reference prices for base billing purposes. B.AI may provide lower actual usage costs through top-up bonuses and account benefits. Specific prices, bonus Credits, and account benefits are subject to the platform display and final billing records.
+:::
diff --git a/docs/llmservice/pricing-and-usage.md b/docs/llmservice/pricing-and-usage.md
@@ -12,20 +12,8 @@ The platform uses a unified Credits system to measure and settle usage across al
 
 **Model pricing:** Different AI models have different pricing based on their capabilities and compute cost. In general, more capable models consume more Credits. Cache-enabled requests may incur separate cache write and cache read usage. Web search incurs an additional per-use charge. Some models do not support web search and are marked with `-`. See the table below for detailed pricing:
 
-:::caution Planned DeepSeek API Pricing Adjustment
-Due to a recent pricing adjustment by DeepSeek, B.AI plans to make a corresponding adjustment to pricing for DeepSeek API services. Please plan your usage accordingly.
-
-The adjustment scope, effective date, and final prices are subject to the formal announcement and platform display.
-:::
-
-:::tip 🎁 Limited-time offer: GLM-5.2 at 40% off
-Offer starts August 12, 2026.
-
-**Eligibility:** This offer applies to GLM-5.2 requests made through the B.AI API and B.AI web app.
-
-For a limited time, eligible requests are billed at 60% of the standard reference price: Input `0.84`, Cache Write `0.84`, Cache Read `0.168`, and Output `2.64` Credits/Token.
-
-The table below continues to show standard reference prices. Offer end time, eligibility, actual settlement price, and final billing are subject to the platform display and final billing records.
+:::info Current promotions and pricing updates
+The table below lists standard reference prices only. For current limited-time offers and pricing updates, see [Promotions and Pricing Updates](./promotions-and-pricing-notices.md). Eligibility, actual settlement prices, and final billing are subject to the platform display.
 :::
 
 | Model             | Input (Credits/Token) | Cache Write (Credits/Token) | Cache Read (Credits/Token) | Output (Credits/Token) | Web Search (Credits/Use) |
@@ -38,11 +26,13 @@ The table below continues to show standard reference prices. Offer end time, eli
 | Qwen3.8-Max       |                  2.00 |                        2.00 |                       0.25 |                   6.00 |                        - |
 | Qwen3.7-Max       |                  1.65 |                        1.65 |                       0.33 |                  4.951 |                        - |
 | Qwen3.6-27B       |                  0.19 |                        0.19 |                      0.019 |                   2.99 |                        - |
+| GLM-5.3           |                  1.40 |                        1.40 |                       0.26 |                   4.40 |                        - |
 | GLM-5.2           |                  1.40 |                        1.40 |                       0.28 |                   4.40 |                        - |
 | GLM-5.1           |                  1.40 |                        1.40 |                       0.28 |                   4.40 |                        - |
 | DeepSeek V3.2     |                  0.29 |                        0.29 |                      0.145 |                   0.44 |                        - |
-| DeepSeek V4 Flash |                  0.28 |                        0.28 |                     0.0056 |                   0.56 |                        - |
+| DeepSeek-V4-Flash |                  0.28 |                        0.28 |                     0.0056 |                   0.56 |                        - |
 | DeepSeek V4 Pro   |                  0.87 |                        0.87 |                     0.0087 |                   1.74 |                        - |
+| Grok 4.6          |                  2.00 |                        2.00 |                       0.50 |                   6.00 |                        - |
 | Grok 4.5          |                  2.00 |                        2.00 |                       0.30 |                   6.00 |                        - |
 | GPT-5.6 Sol       |                  5.00 |                        6.25 |                       0.50 |                  30.00 |                   10,000 |
 | GPT-5.6 Terra     |                  2.00 |                        2.50 |                       0.20 |                  12.00 |                   10,000 |

diff --git a/docs/llmservice/promotions-and-pricing-notices.md b/docs/llmservice/promotions-and-pricing-notices.md
@@ -0,0 +1,40 @@
+# Promotions and Pricing Updates
+
+This page summarizes current limited-time offers and pricing updates. The [Pricing and Usage](./pricing-and-usage.md) table continues to show standard reference prices. Eligibility, offer periods, actual settlement prices, and final billing are subject to the platform display.
+
+## Promotions
+
+:::tip 🔥 Limited-Time Free Access: DeepSeek-V4-Flash
+Offer starts August 17, 2026.
+
+This offer applies to B.AI Chat and API:
+
+* **Chat:** DeepSeek-V4-Flash usage is billed at `0 Credits` during the offer.
+* **API:** Usage is billed at `0 Credits` during the offer. No per-request, input, cache write, cache read, or output token fees apply.
+
+After the offer ends, DeepSeek-V4-Flash will return to standard pricing. See the [model details](./models/deepseek-v4-flash.md).
+:::
+
+:::note 🎁 Limited-time offer: GLM-5.2 at 40% off
+Offer starts August 12, 2026.
+
+**Eligibility:** This offer applies to GLM-5.2 requests made through the B.AI API and B.AI web app.
+
+For a limited time, eligible requests are billed at 60% of the standard reference price: Input `0.84`, Cache Write `0.84`, Cache Read `0.168`, and Output `2.64` Credits/Token. See the [model details](./models/glm-5-2.md).
+:::
+
+:::note 🎁 Limited-time offer: GLM-5.3 at 10% off
+Offer starts August 14, 2026.
+
+**Eligibility:** This offer applies to GLM-5.3 requests made through the B.AI API and B.AI web app.
+
+For a limited time, eligible requests are billed at 90% of the standard reference price: Input `1.26`, Cache Write `1.26`, Cache Read `0.234`, and Output `3.96` Credits/Token. See the [model details](./models/glm-5-3.md).
+:::
+
+## Pricing Updates
+
+:::caution Planned DeepSeek API Pricing Adjustment
+Due to a recent pricing adjustment by DeepSeek, B.AI plans to make a corresponding adjustment to pricing for DeepSeek API services. Please plan your usage accordingly.
+
+The adjustment scope, effective date, and final prices are subject to the formal announcement and platform display.
+:::
diff --git a/i18n/zh-Hans/docusaurus-plugin-content-docs/current.json b/i18n/zh-Hans/docusaurus-plugin-content-docs/current.json
@@ -71,6 +71,10 @@
     "message": "定价与用量",
     "description": "The label for the doc item Pricing and Usage in sidebar docsSidebar"
   },
+  "sidebar.docsSidebar.doc.Promotions and Pricing Updates": {
+    "message": "活动与调整公告",
+    "description": "The label for the doc item Promotions and Pricing Updates in sidebar docsSidebar"
+  },
   "sidebar.docsSidebar.doc.Invitation Rewards": {
     "message": "邀请返利规则",
     "description": "The label for the doc item Invitation Rewards in sidebar docsSidebar"