[Oseltamivir]

Chinese dashboard copy mislabeled modeled facility power as measured power, expressed ratios as ambiguous increases, and obscured date boundaries and parameter explanations. This corrects those meanings, restores precise parallelism and serving terminology, and improves related labels and share text. Calculator comparisons now explicitly say “是…的 N 倍”.

These are presentation-only corrections: filters, selectors, defaults, calculations, API responses and skills behavior are unchanged. The ratio formatter is shared by official and unofficial-run comparison displays. English strings and rendering remain unchanged.

Local smoke testing with all three PRs combined passed: 60 component checks and 129 integration checks, including Chinese desktop/mobile paths and official/unofficial-run overlays. Server-rendered checks also confirmed the repaired multiplication expression, SGLANG_OPT_*, and API fallback descriptions.

Validation details

Ran the repository’s quick component and integration suites plus the fleet lifecycle E2E spec against the combined changes with E2E_FIXTURES=1. The component assertions were updated to follow the corrected 投机解码 copy. HTTP checks returned 200 and verified rendered content on /zh/blog/inferencemax-open-source-inference-benchmarking, /zh/blog/mi355x-deepseek-v4-pro-sglang-110x-in-26-days, and /zh/api. This is automated coverage; it does not replace the outstanding human editorial and visual review.

Chinese copy review

Reviewed against the English siblings with .claude/skills/review-zh-copy/SKILL.md, using separate fidelity and naturalness checks. Disputed substitutions were omitted, including 功率 → 功耗, 编程 → 编码, 事件 → 故障, and 结果格式 → 契约. The video telemetry tag measurement is retained because the English explicitly includes warmup.

A Chinese-speaking maintainer's final editorial review and desktop/mobile visual review remain outstanding.

AI model disclosure

• GPT-6: implementation, source verification and regression tests, through a delegated Codex agent. The runtime did not expose a verifiable exact model/version identifier.
• claude-opus-5-5: advisory review of every changed Chinese passage for fidelity and naturalness, plus follow-up review of corrections.

PR checklist

• AI model disclosure is current.
• Presentation-only copy: no public view, filter, parameter or calculation behavior changes; API and inferencex-skills changes are unnecessary.
• Existing English content is preserved; Chinese siblings are updated.
• Chinese-speaking maintainer has manually reviewed every changed sentence.
• Changed copy has been checked visually at desktop and mobile widths.

中文

本地将三个 PR 的改动合并后运行 smoke 检查，60 项组件检查和 129 项集成检查全部通过，覆盖中文桌面端、移动端及官方/非官方运行叠加。服务端渲染检查也确认了乘法表达式、SGLANG_OPT_* 和 API 回退说明。自动化检查不能替代尚未完成的人工编辑与视觉审核。

修正中文仪表板中将设施级建模功耗误写为实测功耗、倍数关系含糊、日期边界不清晰等问题，并改善相关控件、分离式推理说明、并行度提示与分享文案。计算器对比现在使用“是…的 N 倍”，明确表示比值。

本次仅调整展示文案，不改变筛选器、默认值、计算、API 响应或 skills 行为；官方运行和非官方运行共用修正后的倍数文案。英文文案与渲染保持不变。未采纳有争议的机械替换，例如统一将“功率”改成“功耗”，或将“事件”改成“故障”。仍需中文维护者逐句审核，以及桌面端和移动端视觉检查。

AI 使用说明：委派的 Codex agent 使用 GPT-6 进行实现、源码核对与回归测试；运行环境未提供可核实的精确模型版本。claude-opus-5-5 根据仓库的 review-zh-copy skill 对全部改动进行语义准确性与中文自然度审核，并复核后续修正。

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
inferencemax-app	Ready	Preview	Oct 25, 2026 9:45am UTC

[cursor]

Bugbot is paused — your team has an unpaid invoice

Bugbot is paused because your team has an unpaid invoice.

Pay it from the billing page in the Cursor dashboard and Bugbot resumes automatically.

(requestId: 42849fa0-0390-4675-b500-4a23274ed439)

[cursor]

Bugbot is paused — your team has an unpaid invoice

Bugbot is paused because your team has an unpaid invoice.

Pay it from the billing page in the Cursor dashboard and Bugbot resumes automatically.

(requestId: e1955961-a165-410f-b673-85be69b1fec4)