From 4dcb37d6ebab35a96aaabd6414496cb006ce6024 Mon Sep 17 00:00:00 2001 From: sophia Date: Sat, 13 Jun 2026 11:59:21 +0800 Subject: [PATCH 1/2] Remove deprecated LLM model docs --- docs/llmservice/models/claude-fable-5.md | 53 ------------------- docs/llmservice/models/kimi-k2.6.md | 50 ----------------- docs/llmservice/pricing-and-usage.md | 2 - .../llmservice/models/claude-fable-5.md | 53 ------------------- .../current/llmservice/models/kimi-k2.6.md | 50 ----------------- .../current/llmservice/pricing-and-usage.md | 2 - .../current/sidebars.js | 2 - sidebars.js | 2 - 8 files changed, 214 deletions(-) delete mode 100644 docs/llmservice/models/claude-fable-5.md delete mode 100644 docs/llmservice/models/kimi-k2.6.md delete mode 100644 i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/claude-fable-5.md delete mode 100644 i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/kimi-k2.6.md diff --git a/docs/llmservice/models/claude-fable-5.md b/docs/llmservice/models/claude-fable-5.md deleted file mode 100644 index 09e98d52..00000000 --- a/docs/llmservice/models/claude-fable-5.md +++ /dev/null @@ -1,53 +0,0 @@ -# Claude Fable 5 - -## Overview - -Claude Fable 5 is a high-capability Anthropic model available on B.AI for advanced reasoning, coding, long-context analysis, and agentic workflows. It is designed for complex tasks that require sustained context, tool-assisted execution, and high-quality structured outputs. Specific capabilities, context limits, tool support, and availability may vary by B.AI model catalog and platform configuration. - -## Key Features - -* **Advanced Reasoning**: Suitable for complex analytical, technical, and professional knowledge tasks. -* **Software Engineering Workflows**: Designed for coding assistance, debugging, refactoring, code review, and multi-step implementation planning. -* **Long-Context Tasks**: Supports extended analysis across large codebases, long documents, and multi-turn work sessions when enabled by the platform configuration. -* **Agentic and Tool-Assisted Workflows**: Suitable for workflows that rely on tool use, function calling, code execution, MCP, or compatible agent environments. -* **Multimodal Understanding**: Supports text and image input for document, screenshot, chart, and diagram understanding where available. - -## Best Use Cases - -* **Complex Software Engineering**: Large feature work, repository-scale refactors, migration planning, bug investigation, and code review. -* **Extended Agentic Workflows**: Multi-step tasks that require planning, tool use, verification, and sustained context over longer sessions. -* **Research and Knowledge Work**: Analysis and synthesis across technical documents, legal or financial materials, and structured research sources. -* **Visual Document Analysis**: Understanding screenshots, diagrams, charts, PDFs, and other image-based materials when supported by the workflow. - -## Capabilities and Limitations - -| Capability | Description | -| :----------------- | :-------------------------------------------------------------------------------------------------- | -| **Reasoning** | Advanced reasoning for complex professional and technical tasks | -| **Coding** | Strong coding, debugging, refactoring, and code review capabilities | -| **Agentic** | Suitable for long-running tool workflows and multi-step agent tasks | -| **Computer Use** | Can support browser and desktop interaction through compatible tools and environments | -| **Multimodal** | Text and image input; text output | -| **Context Window** | Up to 1,000,000 tokens, subject to platform configuration | -| **Max Output** | Up to 128,000 tokens, subject to platform configuration | -| **Tool Use** | Function calling, code execution, MCP support, adaptive thinking, and compatible agent workflows | -| **Multilingual** | Strong multilingual performance across major world languages | - -### Known Limitations - -* Specific capability availability may depend on the B.AI integration, Anthropic platform support, plan settings, and rollout status. -* Web access, code execution, computer use, and external actions require compatible tools or integrations. -* Image input is supported, but native audio or video input is not listed for this model. -* Public evaluations, third-party comparisons, policy behavior, and implementation details may change over time, so they are not treated as fixed guarantees in this documentation. - -## Credits Usage - -| Model | Input (Credits/Token) | Cache Write (Credits/Token) | Cache Read (Credits/Token) | Output (Credits/Token) | Web Search (Credits/Use) | Billing Notes | -| :--- | --------------------: | --------------------------: | -------------------------: | ---------------------: | -----------------------: | :--- | -| **Claude Fable 5** | `10.00` | `12.50` | `1.00` | `50.00` | `10,000` | - | - -:::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. -::: - -* **Prompt caching**: Cache writes are charged at 1.25x base input price for the 5-minute TTL option, or 2x base input price for the 1-hour TTL option. Cache reads are charged at 0.1x base input price. Prompt caching requires a minimum of 1,024 tokens. diff --git a/docs/llmservice/models/kimi-k2.6.md b/docs/llmservice/models/kimi-k2.6.md deleted file mode 100644 index 9d73ecc0..00000000 --- a/docs/llmservice/models/kimi-k2.6.md +++ /dev/null @@ -1,50 +0,0 @@ -# Kimi K2.6 - -## Overview - -Kimi K2.6 is Moonshot AI's open-weight multimodal model, released on April 20, 2026. It is the third K2-class model in nine months, following K2 and K2.5. Built on a 1-trillion-parameter Mixture-of-Experts architecture with 32 billion active parameters per token, K2.6 combines native multimodal input, advanced agent swarm orchestration, and strong coding performance. - -## Key Features - -* **Native Multimodal Architecture**: Supports text, image, and video input through the custom MoonViT vision encoder. Video input is new in K2.6 and supports mp4, mov, avi, and webm formats. -* **Agent Swarm Orchestration**: Supports up to 300 concurrent sub-agents per task and 4,000 coordinated steps, with a 96.6% tool-invocation success rate, up from 91% on K2.5. -* **Coding Performance**: Achieves SWE-Bench Pro 58.6%, SWE-bench Verified 80.2%, LiveCodeBench v6 89.6%, and Terminal-Bench 2.0 66.7%. -* **Modified MIT License**: Open weights are available on Hugging Face and are free for commercial use below 100M MAU or $20M monthly revenue. - -## Best Use Cases - -* **End-to-End Coding & UI Generation**: Well suited for transforming prompts and visual inputs into production-ready interfaces and lightweight full-stack workflows across Python, Rust, and Go. -* **Multi-Agent Systems**: The 300-agent swarm capacity with 4,000-step coordination makes it effective for complex autonomous workflows that require long-context stability. -* **Cost-Effective Multimodal Processing**: Offers strong multimodal and agentic performance at a lower cost than many proprietary multimodal alternatives. - -## Capabilities and Limitations - -| Capability | Description | -| :----------------- | :------------------------------------------------------------------------------------------------------------ | -| **Reasoning** | AIME 2026: 96.4%, GPQA-Diamond: 90.5%, HLE with tools: 54.0% | -| **Coding** | SWE-Bench Pro 58.6%, SWE-bench Verified 80.2%, LiveCodeBench v6 89.6%, Terminal-Bench 2.0 66.7% | -| **Multimodal** | Text, image (png, jpeg, webp, gif), and video (mp4, mov, avi, webm) input through the MoonViT vision encoder | -| **Response Speed** | Optimized for throughput in agentic workflows; specific tokens-per-second metrics vary by deployment | -| **Context Window** | 262K tokens | -| **Max Output** | 16K tokens, up to 98K in extended mode | -| **Tool Use** | 96.6% tool-invocation success, 4,000+ tool calls per session, and multi-agent handoffs | -| **Multilingual** | 160K vocabulary optimized for code and non-English text; SWE-bench Multilingual 76.7% | - -### Known Limitations - -* Multimodal benchmark performance is weaker than top proprietary models on some vision tasks such as MMMU-Pro and MathVision. -* URL-based image input is not supported through the API; only base64-encoded content or file upload is supported. -* Image resolution is capped at 4K, video at 2K, and the full request body must remain under 100MB. -* Pure math reasoning trails some higher-end proprietary models on benchmarks such as AIME 2026 and GPQA-Diamond. -* The 262K context window is smaller than some proprietary alternatives offering 1M+ tokens. -* Independent reviews note only marginal improvement over K2.5 on day-to-day tasks and weaker performance on some domain-specific workloads. - -## Credits Usage - -| Model | Input (Credits/Token) | Cache Write (Credits/Token) | Cache Read (Credits/Token) | Output (Credits/Token) | Web Search (Credits/Use) | Billing Notes | -| :--- | --------------------: | --------------------------: | -------------------------: | ---------------------: | -----------------------: | :--- | -| **Kimi K2.6** | `0.95` | `0.95` | `0.16` | `4.00` | `-` | - | - -:::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 index ed0ba0b7..0197ebc3 100644 --- a/docs/llmservice/pricing-and-usage.md +++ b/docs/llmservice/pricing-and-usage.md @@ -16,7 +16,6 @@ The platform uses a unified Credits system to measure and settle usage across al | :---------------- | --------------------: | --------------------------: | -------------------------: | ---------------------: | -----------------------: | | MiniMax M3 | 0.30 | 0.30 | 0.06 | 1.20 | - | | MiniMax M2.7 | 0.30 | 0.375 | 0.06 | 1.20 | - | -| Kimi K2.6 | 0.95 | 0.95 | 0.16 | 4.00 | - | | Kimi K2.5 | 0.59 | 0.59 | 0.177 | 3.00 | - | | Qwen3.6-27B | 0.19 | 0.19 | 0.19 | 2.99 | - | | GLM-5.1 | 1.40 | 1.40 | 0.26 | 4.40 | - | @@ -33,7 +32,6 @@ The platform uses a unified Credits system to measure and settle usage across al | GPT-5 Mini | 0.25 | 0.25 | 0.025 | 2.00 | 10,000 | | GPT-5.4 Nano | 0.20 | 0.20 | 0.02 | 1.25 | 10,000 | | GPT-5 Nano | 0.05 | 0.05 | 0.005 | 0.40 | - | -| Claude Fable 5 | 10.00 | 12.50 | 1.00 | 50.00 | 10,000 | | Claude Opus 4.8 | 5.00 | 6.25 | 0.50 | 25.00 | 10,000 | | Claude Opus 4.7 | 5.00 | 6.25 | 0.50 | 25.00 | 10,000 | | Claude Opus 4.6 | 5.00 | 6.25 | 0.50 | 25.00 | 10,000 | diff --git a/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/claude-fable-5.md b/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/claude-fable-5.md deleted file mode 100644 index c355efe5..00000000 --- a/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/claude-fable-5.md +++ /dev/null @@ -1,53 +0,0 @@ -# Claude Fable 5 - -## 概述 - -Claude Fable 5 是 B.AI 上可用的 Anthropic 高能力模型，面向复杂推理、代码任务、长上下文分析和 Agent 工作流。它适合需要持续上下文、工具辅助执行和高质量结构化输出的复杂任务。具体能力、上下文长度、工具支持和可用状态可能会随 B.AI 模型目录和平台配置调整。 - -## 核心特性 - -* **高级推理能力**：适合复杂分析、技术任务和专业知识工作。 -* **软件工程工作流**：面向代码辅助、调试、重构、代码审查和多步骤实现规划。 -* **长上下文任务**：在平台配置支持时，可用于大型代码库、长文档和多轮工作会话的持续分析。 -* **Agent 与工具辅助工作流**：适合依赖工具调用、函数调用、代码执行、MCP 或兼容 Agent 环境的工作流。 -* **多模态理解**：在可用场景下，支持文本和图像输入，可用于文档、截图、图表和技术示意图理解。 - -## 适用场景 - -* **复杂软件工程**：大型功能开发、仓库级重构、迁移规划、Bug 排查和代码审查。 -* **长时间 Agent 工作流**：需要规划、工具调用、验证和持续上下文保持的多步骤任务。 -* **研究与知识工作**：技术文档、法律或金融材料以及结构化研究资料的分析与综合。 -* **视觉文档分析**：在工作流支持时，可处理截图、图表、PDF 和其他图像型材料。 - -## 能力与限制 - -| 能力维度 | 说明 | -| :--- | :--- | -| **推理能力** | 适合复杂专业任务和技术任务的高级推理 | -| **编程能力** | 具备较强的编码、调试、重构和代码审查能力 | -| **Agent 能力** | 适合长时间工具调用工作流和多步骤 Agent 任务 | -| **计算机操作** | 可通过兼容工具和环境支持浏览器及桌面交互 | -| **多模态能力** | 支持文本和图像输入；输出为文本 | -| **上下文窗口** | 最高 1,000,000 tokens，具体以平台配置为准 | -| **最大输出** | 最高 128,000 tokens，具体以平台配置为准 | -| **工具调用** | 支持函数调用、代码执行、MCP、自适应思考和兼容 Agent 工作流 | -| **多语言能力** | 在主要世界语言上具备较强的多语言表现 | - -### 已知限制 - -* 具体能力可用性可能取决于 B.AI 集成、Anthropic 平台支持、套餐配置和功能上线状态。 -* 联网访问、代码执行、计算机操作和外部动作需要兼容工具或集成支持。 -* 支持图像输入，但该模型未标明原生音频或视频输入能力。 -* 公开评测、第三方对比、策略行为和实现细节可能随时间变化，因此本文档不将其作为固定承诺。 - -## 积分消耗 - -| 模型名称 | 输入 (Credits/Token) | Cache Write (Credits/Token) | Cache Read (Credits/Token) | 输出 (Credits/Token) | 网页搜索（Credits/次） | 计费说明 | -| :--- | --------------------: | --------------------------: | -------------------------: | -------------------: | ---------------------: | :--- | -| **Claude Fable 5** | `10.00` | `12.50` | `1.00` | `50.00` | `10,000` | - | - -:::info 价格说明 -文档价格为 B.AI 平台模型标准参考价，仅供基础计费说明使用。B.AI 可能会通过充值赠送及账户权益等方式，为用户提供更低的实际使用成本。具体价格、赠送积分及账户权益请以平台页面展示及最终账单为准。 -::: - -* **Prompt caching**：缓存写入按基础输入价格的 1.25x 计费（5 分钟 TTL），或按基础输入价格的 2x 计费（1 小时 TTL）。缓存读取按基础输入价格的 0.1x 计费。使用 Prompt caching 时，最低需要 1,024 tokens。 diff --git a/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/kimi-k2.6.md b/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/kimi-k2.6.md deleted file mode 100644 index b1f75aac..00000000 --- a/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/kimi-k2.6.md +++ /dev/null @@ -1,50 +0,0 @@ -# Kimi K2.6 - -## 概述 - -Kimi K2.6 是 Moonshot AI 于 2026 年 4 月 20 日发布的开放权重多模态模型，也是 K2 和 K2.5 之后，九个月内推出的第三个 K2 系列模型。该模型采用 1 万亿参数的 Mixture-of-Experts 架构，每个 token 激活约 320 亿参数，结合了原生多模态输入、先进的 Agent swarm 编排能力，以及较强的编程表现。 - -## 核心特性 - -* **原生多模态架构**：通过自研 MoonViT 视觉编码器支持文本、图像和视频输入。K2.6 新增视频输入能力，支持 mp4、mov、avi 和 webm 格式。 -* **Agent Swarm 编排**：单个任务最多支持 300 个并发子 Agent 和 4,000 个协同步骤，工具调用成功率达到 96.6%，高于 K2.5 的 91%。 -* **编程能力**：SWE-Bench Pro 58.6%，SWE-bench Verified 80.2%，LiveCodeBench v6 89.6%，Terminal-Bench 2.0 66.7%。 -* **修改版 MIT 许可**：开放权重已发布在 Hugging Face 上，当月活低于 1 亿或月营收低于 2,000 万美元时，可免费用于商业场景。 - -## 适用场景 - -* **端到端编码与 UI 生成**：适合将文本提示和视觉输入转化为可落地的界面，以及轻量级全栈工作流，覆盖 Python、Rust 和 Go 等语言。 -* **多 Agent 系统**：300 Agent 并发能力和 4,000 步协同上限，使其适合需要长上下文稳定性的复杂自主工作流。 -* **高性价比多模态处理**：在较低成本下提供多模态与 Agent 工作流能力，适合对成本敏感的高频任务场景。 - -## 能力与限制 - -| 能力维度 | 说明 | -| :--- | :--- | -| **推理能力** | AIME 2026：96.4%，GPQA-Diamond：90.5%，HLE with tools：54.0% | -| **编程能力** | SWE-Bench Pro 58.6%，SWE-bench Verified 80.2%，LiveCodeBench v6 89.6%，Terminal-Bench 2.0 66.7% | -| **多模态能力** | 通过 MoonViT 视觉编码器支持文本、图像（png、jpeg、webp、gif）和视频（mp4、mov、avi、webm）输入 | -| **响应速度** | 面向 Agent 工作流做了吞吐优化，具体 tokens/s 表现会随部署环境变化 | -| **上下文窗口** | 262K tokens | -| **最大输出** | 16K tokens，扩展模式下最高可达 98K | -| **工具调用** | 工具调用成功率 96.6%，单次会话支持 4,000+ 次工具调用和多 Agent 交接 | -| **多语言能力** | 160K 词表针对代码和非英语文本做了优化；SWE-bench Multilingual 76.7% | - -### 已知限制 - -* 在部分视觉基准上，多模态表现仍弱于顶级专有模型，例如 MMMU-Pro 和 MathVision。 -* API 不支持通过 URL 直接传入图片，只支持 base64 编码内容或文件上传。 -* 图片分辨率上限为 4K，视频分辨率上限为 2K，整个请求体需控制在 100MB 以内。 -* 在纯数学推理任务上，AIME 2026 和 GPQA-Diamond 等基准仍落后于部分更高端的专有模型。 -* 262K 上下文窗口小于部分提供 1M+ tokens 的专有替代方案。 -* 独立评测认为其相较 K2.5 的日常任务提升有限，在某些垂直领域任务上仍存在短板。 - -## 积分消耗 - -| 模型名称 | 输入 (Credits/Token) | Cache Write (Credits/Token) | Cache Read (Credits/Token) | 输出 (Credits/Token) | 网页搜索（Credits/次） | 计费说明 | -| :--- | --------------------: | --------------------------: | -------------------------: | -------------------: | ---------------------: | :--- | -| **Kimi K2.6** | `0.95` | `0.95` | `0.16` | `4.00` | `-` | - | - -:::info 价格说明 -文档价格为 B.AI 平台模型标准参考价，仅供基础计费说明使用。B.AI 可能会通过充值赠送及账户权益等方式，为用户提供更低的实际使用成本。具体价格、赠送积分及账户权益请以平台页面展示及最终账单为准。 -::: diff --git a/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/pricing-and-usage.md b/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/pricing-and-usage.md index 66ccf810..94e87751 100644 --- a/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/pricing-and-usage.md +++ b/i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/pricing-and-usage.md @@ -16,7 +16,6 @@ | :--- | --------------------: | ------------------------: | ------------------------: | --------------------: | ---------------------: | | MiniMax M3 | 0.30 | 0.30 | 0.06 | 1.20 | - | | MiniMax M2.7 | 0.30 | 0.375 | 0.06 | 1.20 | - | -| Kimi K2.6 | 0.95 | 0.95 | 0.16 | 4.00 | - | | Kimi K2.5 | 0.59 | 0.59 | 0.177 | 3.00 | - | | Qwen3.6-27B | 0.19 | 0.19 | 0.19 | 2.99 | - | | GLM-5.1 | 1.40 | 1.40 | 0.26 | 4.40 | - | @@ -33,7 +32,6 @@ | GPT-5 Mini | 0.25 | 0.25 | 0.025 | 2.00 | 10,000 | | GPT-5.4 Nano | 0.20 | 0.20 | 0.02 | 1.25 | 10,000 | | GPT-5 Nano | 0.05 | 0.05 | 0.005 | 0.40 | - | -| Claude Fable 5 | 10.00 | 12.50 | 1.00 | 50.00 | 10,000 | | Claude Opus 4.8 | 5.00 | 6.25 | 0.50 | 25.00 | 10,000 | | Claude Opus 4.7 | 5.00 | 6.25 | 0.50 | 25.00 | 10,000 | | Claude Opus 4.6 | 5.00 | 6.25 | 0.50 | 25.00 | 10,000 | diff --git a/i18n/zh-Hans/docusaurus-plugin-content-docs/current/sidebars.js b/i18n/zh-Hans/docusaurus-plugin-content-docs/current/sidebars.js index 82c780c8..e0985e35 100644 --- a/i18n/zh-Hans/docusaurus-plugin-content-docs/current/sidebars.js +++ b/i18n/zh-Hans/docusaurus-plugin-content-docs/current/sidebars.js @@ -229,7 +229,6 @@ const sidebars = { 'llmservice/models/claude-opus-4-6', 'llmservice/models/claude-opus-4-7', 'llmservice/models/claude-opus-4-8', - 'llmservice/models/claude-fable-5', 'llmservice/models/claude-sonnet-4-5', 'llmservice/models/claude-sonnet-4-6', 'llmservice/models/deepseek-v3.2', @@ -240,7 +239,6 @@ const sidebars = { 'llmservice/models/gemini-3-flash', 'llmservice/models/glm-5-1', 'llmservice/models/glm-5', - 'llmservice/models/kimi-k2.6', 'llmservice/models/kimi-k2.5', 'llmservice/models/qwen3.6-27b', 'llmservice/models/minimax-m3', diff --git a/sidebars.js b/sidebars.js index 5e63431e..dc1aade3 100644 --- a/sidebars.js +++ b/sidebars.js @@ -226,7 +226,6 @@ const sidebars = { 'llmservice/models/claude-opus-4-6', 'llmservice/models/claude-opus-4-7', 'llmservice/models/claude-opus-4-8', - 'llmservice/models/claude-fable-5', 'llmservice/models/claude-sonnet-4-5', 'llmservice/models/claude-sonnet-4-6', 'llmservice/models/deepseek-v3.2', @@ -237,7 +236,6 @@ const sidebars = { 'llmservice/models/gemini-3-flash', 'llmservice/models/glm-5-1', 'llmservice/models/glm-5', - 'llmservice/models/kimi-k2.6', 'llmservice/models/kimi-k2.5', 'llmservice/models/qwen3.6-27b', 'llmservice/models/minimax-m3', From e2c9e42d243a600bde4a41a72530d224a4c0abdd Mon Sep 17 00:00:00 2001 From: jerryji-prog Date: Mon, 15 Jun 2026 17:11:24 +0800 Subject: [PATCH 2/2] add version --- package.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/package.json b/package.json index 448c9505..c2a71424 100644 --- a/package.json +++ b/package.json @@ -1,6 +1,6 @@ { "name": "@x402-tron/docs", - "version": "1.2.26", + "version": "1.2.27", "description": "x402-tron documentation", "license": "MIT", "scripts": {