Conversation
Code Review ReportProject: Bank of AI / B.AI Documentation PR OverviewBranch Information
Commit History
Review SummaryVerdict
Findings at a Glance
SummaryThis PR makes three categories of changes to documentation files: (1) a platform branding rename from "Bank of AI" to "B.AI" across all English and Chinese model docs, (2) a uniform ~10% price increase for all five Claude models (Haiku 4.5, Sonnet 4.5, Sonnet 4.6, Opus 4.5, Opus 4.6), and (3) additions of two new models to the pricing table (DeepSeek V3.2 and GPT-5.4 Pro) alongside minor naming normalization (hyphenated GPT model names changed to space-separated). The overall intent of the PR is clear and the pricing numbers are internally consistent across all 12 files. However, there is one Major defect: the Chinese-language localization files all contain a spurious Chinese character Change Summary1. Platform Brand Rename (All Model Docs — EN & ZH)
Purpose: Rebrand platform references from the full name "Bank of AI" to the shorter abbreviated form "B.AI". 2. Claude Model Pricing Increase (~10% across all tiers)
Purpose: Reflect a uniform ~10% price increase for all Claude models on the platform. 3. Pricing Table Additions & Normalization
Purpose: Expand the pricing table with two newly supported models and standardize model name formatting. Detailed FindingsMajor[MJ-01] Garbled Brand Name in All Chinese-Language Model Docs — "白B.AI"
Description
Code -Claude Haiku 4.5 是 Anthropic 推出的轻量高速模型，已集成至 Bank of AI 平台。
+Claude Haiku 4.5 是 Anthropic 推出的轻量高速模型，已集成至 白B.AI 平台。
# ^^^
# Extraneous character "白" (white) must be removed(Same pattern in Recommendation -已集成至 白B.AI 平台
+已集成至 B.AI 平台All four zh-Hans model files must be corrected before merge. Minor[MN-01] GPT-5.4 Pro Web Search Pricing Missing Without Explanation
Description
Code | GPT-5.4 Pro | 30.00 | 180.00 | - |Recommendation
> **Note:** `-` in the Web Search column indicates the feature is not currently supported for that model.[MN-02] No Dedicated Model Documentation Page for New Models (DeepSeek V3.2, GPT-5.4 Pro)
Description
Recommendation
[MN-03] Missing H1 Heading Not Backfilled in
|
| Property | Value |
|---|---|
| Severity | Minor |
| Category | Documentation / Consistency |
| File | i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/claude-sonnet-4-6.md |
Description
The English
pricing-and-usage.mdand its zh-Hans counterpart both received a new top-level H1 heading (# Pricing and Usage/# 定价与用量) in this PR. However, theclaude-sonnet-4-6.mdzh-Hans file received only a pricing change — unlike its zh-Hans siblings (haiku, opus-4-5, opus-4-6, sonnet-4-5) which also received the brand rename. Looking at the diff,i18n/zh-Hans/.../claude-sonnet-4-6.mdonly changes the pricing row. If the model overview paragraph in that file still contains the old "Bank of AI" phrasing (not changed in this PR's diff), the rename is incomplete for that file.
Recommendation
Verify whether
i18n/zh-Hans/.../claude-sonnet-4-6.mdcontains "Bank of AI" or an older platform name in its overview section, and apply the brand rename if so. The Englishclaude-sonnet-4-6.mdwas similarly only updated for pricing (not branding), suggesting the overview text may already use a different phrasing — confirm this is intentional.
Suggestions
[S-01] No Changelog or Release Note for Pricing Increase
File: docs/llmservice/pricing-and-usage.md
Description: The ~10% Claude pricing increase is a significant user-facing change. There is no changelog entry, announcement banner, or "effective date" metadata in any of the updated files. Users who have bookmarked individual model pages may miss the price change.
Suggestion: Add an > **Pricing updated:** YYYY-MM-DD callout block at the top of each model pricing section, or maintain a CHANGELOG.md / pricing-history.md that records pricing changes with effective dates.
[S-02] GPT Model Name Normalization is Partial
File: docs/llmservice/pricing-and-usage.md
Description: This PR renames GPT-5.4-mini → GPT-5.4 Mini, GPT-5-mini → GPT-5 Mini, etc. (hyphen-separated to space-and-capitalized). However, GLM-5 and MiniMax M2.5 retain their original formats. If the intent is a site-wide naming convention change, the normalization should be consistently applied to all model names in the table.
Suggestion: Define an explicit naming convention (e.g., <Brand> <Version> <Tier>) and apply it uniformly, or document exceptions for models whose official names use hyphens.
Positive Observations
| Area | Observation |
|---|---|
| Consistency | Pricing changes are perfectly consistent across all 12 files (EN and zh-Hans model pages + both pricing tables). No mismatch between individual model pages and the central pricing table. |
| Bilingual parity | Every structural change to pricing-and-usage.md (new H1, new model rows, name normalizations) is mirrored exactly in the zh-Hans version. |
| Numerical accuracy | The 10% increase is applied uniformly: Haiku 1.00→1.10/5.00→5.50, Sonnet 3.00→3.30/15.00→16.50, Opus 5.00→5.50/25.00→27.50. All ratios are preserved. |
| Table alignment | The pricing table column alignment (right-aligned numbers) is maintained consistently in both EN and ZH versions. |
| Scope discipline | The PR stays focused — only pricing and branding changes, no unrelated edits. |
Checklist Results
| Category | Items Checked | Pass | Fail | N/A | Notes |
|---|---|---|---|---|---|
| Correctness | 3 | 2 | 1 | 5 | "白B.AI" brand name error in zh-Hans files |
| Security | 8 | 0 | 0 | 8 | N/A — documentation only |
| Performance | 7 | 0 | 0 | 7 | N/A — documentation only |
| Code Quality | 6 | 4 | 0 | 2 | Naming mostly consistent; minor normalization gap |
| Testing | 7 | 0 | 0 | 7 | N/A — no executable code changed |
| Documentation | 6 | 4 | 2 | 0 | New models lack dedicated pages; no changelog for pricing |
| Compatibility | 5 | 5 | 0 | 0 | Additive changes only; no breaking removals |
| Observability | 4 | 0 | 0 | 4 | N/A — documentation only |
Disclaimer
This is an automated code review. It supplements but does not replace human review. The reviewer analyzed only the diff between main and ai-bankofai-patch-1. Runtime behavior, integration testing, and deployment impact are not covered.
Report generated by Code Review Skill v1.0.0
Date: 2026-04-21
Code Review ReportProject: B.AI Platform Documentation ( PR OverviewBranch Information
Commit History
Review SummaryVerdict
Findings at a Glance
SummaryThis PR is a documentation-only update targeting the B.AI Platform docs site. It covers three areas: (1) a brand rename from "Bank of AI" to "B.AI" across all English and Chinese model pages, (2) a uniform ~10% price increase across all five Claude models in both individual model pages and the shared pricing table, and (3) additions to the pricing table (two new models: DeepSeek V3.2 and GPT-5.4 Pro) along with normalising model display names from hyphenated to space-separated. The changes are straightforward and systematic, but one Major defect was found: the Chinese-language localisation files introduce a spurious CJK character Beyond that defect, three Minor issues relate to inconsistent naming conventions, a new model (DeepSeek V3.2) lacking a dedicated documentation page, and an undocumented GPT-5.4 Pro feature restriction (no web search). Two additional suggestions are offered for long-term quality. Change Summary1. Package Version Bump
Purpose: Reflects the content changes introduced by this PR in a semver-compatible patch release. 2. Brand Rename — English Docs
Purpose: Aligns the English documentation with the updated platform brand identity. 3. Brand Rename — Chinese (zh-Hans) Docs
Purpose: Intended to apply the same brand rename to the Chinese localisation, but introduces a stray 4. Claude Model Price Increases
Purpose: Applies a uniform ~10% credit price increase to all Claude models, synchronised across individual model pages and the master pricing table in both locales. 5. Pricing Table Overhaul
Purpose: Keeps the master pricing reference table current with new models and updated Claude pricing. Also improves document structure by adding a top-level H1 heading and standardises model name formatting. Detailed FindingsMajor[MJ-01] Spurious
|
| Property | Value |
|---|---|
| Severity | Major |
| Category | Documentation / Correctness |
| Files | i18n/zh-Hans/docusaurus-plugin-content-docs/current/llmservice/models/claude-haiku-4-5.md : Line 5 · claude-opus-4-5.md : Line 5 · claude-opus-4-6.md : Line 5 · claude-sonnet-4-5.md : Line 5 |
Description
Every Chinese-locale model file that received a brand rename now reads "白B.AI 平台" (literally "White B.AI Platform") instead of the intended "B.AI 平台". The stray
白character (U+767D) appears to be an accidental IME (Input Method Editor) artefact — the author likely began typing "白" while in Chinese input mode before switching to type the brand name in Latin characters. The resulting string is a live branding error visible to all Simplified Chinese users of the documentation site.
Code (representative hunk from claude-haiku-4-5.md)
-Claude Haiku 4.5 是 Anthropic 推出的轻量高速模型，已集成至 Bank of AI 平台。
+Claude Haiku 4.5 是 Anthropic 推出的轻量高速模型，已集成至 白B.AI 平台。The same pattern appears identically in claude-opus-4-5.md, claude-opus-4-6.md, and claude-sonnet-4-5.md.
Recommendation
Remove the 白 character so the replacement reads B.AI 平台 / B.AI 生态:
-Claude Haiku 4.5 是 Anthropic 推出的轻量高速模型，已集成至 白B.AI 平台。
+Claude Haiku 4.5 是 Anthropic 推出的轻量高速模型，已集成至 B.AI 平台。Apply the same correction to the other three affected Chinese model files.
Minor
[MN-01] Inconsistent Model Name Normalisation — GLM-5 Retains Hyphen
| Property | Value |
|---|---|
| Severity | Minor |
| Category | Code Quality / Consistency |
| File | docs/llmservice/pricing-and-usage.md : Line 17 · i18n/zh-Hans/.../pricing-and-usage.md : Line 17 |
Description
This PR normalises hyphenated model names to space-separated names (e.g.
MiniMax-M2.5→MiniMax M2.5,Kimi-K2.5→Kimi K2.5,GPT-5.4-mini→GPT-5.4 Mini) in both the English and Chinese pricing tables. However,GLM-5was left unchanged and still uses a hyphen. This creates an inconsistent naming convention within the same table.
Code
| MiniMax M2.5 | 0.30 | 1.20 | - | ← hyphen removed ✓
| Kimi K2.5 | 0.23 | 3.00 | - | ← hyphen removed ✓
| GLM-5 | 0.30 | 2.55 | - | ← hyphen still present ✗
| DeepSeek V3.2 | 0.27 | 0.42 | - | ← new entry, no hyphen ✓Recommendation
Either rename
GLM-5→GLM 5to match the new convention, or — if the official model branding uses a hyphen — document that exception and apply the same exception consistently to any other models whose official names require it.
[MN-02] New Model DeepSeek V3.2 Has No Dedicated Documentation Page
| Property | Value |
|---|---|
| Severity | Minor |
| Category | Documentation |
| File | docs/llmservice/pricing-and-usage.md : Line 18 · i18n/zh-Hans/.../pricing-and-usage.md : Line 18 |
Description
DeepSeek V3.2is added to the pricing table in both locales, but this PR does not add a corresponding individual model page underdocs/llmservice/models/(or its Chinese equivalent). Every other model in the pricing table has a dedicated page. Users who click on the model name or navigate to the models section will find no details about DeepSeek V3.2's capabilities, use cases, or context window.
Recommendation
Add
docs/llmservice/models/deepseek-v3-2.md(andi18n/zh-Hans/.../deepseek-v3-2.md) following the same structure as the existing model pages, covering: Overview, Key Features, Use Cases, Capabilities & Limitations, and the pricing table.
[MN-03] GPT-5.4 Pro Missing Web Search — Undocumented Restriction
| Property | Value |
|---|---|
| Severity | Minor |
| Category | Documentation |
| File | docs/llmservice/pricing-and-usage.md : Line 20 · i18n/zh-Hans/.../pricing-and-usage.md : Line 20 |
Description
The newly added
GPT-5.4 Prorow shows-for Web Search cost, unlike the standardGPT-5.4which supports web search at 10,000 Credits/Use. Users would reasonably expect the "Pro" variant to be a superset of the base model's features, so the absence of web search is unexpected and unexplained. The existing footnote in the pricing doc states: "Some models do not support web search and are marked with '-'." — but it gives no further context about why any specific model lacks support.
Recommendation
If GPT-5.4 Pro genuinely does not support web search, add a brief clarifying note below the table — or in the GPT-5.4 Pro model page — explaining this limitation (e.g., due to API-level restrictions or the model being optimised for other workloads). If web search support will be added later, consider noting it as "Coming Soon."
Suggestions
[S-01] No Changelog Entry for the Price Increase
File: package.json (version bump only — no CHANGELOG.md updated)
Description: The version was incremented from 1.2.10 → 1.2.11, but there is no changelog or release notes file tracking what changed. Price increases directly affect platform users' credit spend, making this particularly important to record.
Suggestion: Add a CHANGELOG.md (or equivalent) entry for v1.2.11 summarising: brand rename, ~10% Claude price increase, addition of DeepSeek V3.2 and GPT-5.4 Pro to the catalogue, and model name formatting standardisation.
[S-02] Consider a Pricing Data Source-of-Truth
File: All model .md files + both pricing-and-usage.md files
Description: Prices are currently duplicated across: one individual English model page, one individual Chinese model page, and two pricing table files — for each model. This PR correctly updated all locations in sync, but the risk of inconsistency grows with every future price change. A future contributor editing only the pricing table but forgetting the model page (or vice-versa) would create a discrepancy visible to users.
Suggestion: Explore storing pricing data in a single JSON/YAML data file and using a Docusaurus custom component or MDX import to render it in both the individual model pages and the master pricing table. This would make future price updates a one-line change with guaranteed consistency.
Positive Observations
| Area | Observation |
|---|---|
| Price Consistency | All five Claude model prices were updated in perfect synchronisation across individual English pages, individual Chinese pages, and both pricing tables — no drift detected. |
| Uniform Increase | The ~10% price increase is applied uniformly across all Claude models (Haiku, Sonnet, Opus), making the change easy to audit and reason about. |
| Document Structure | Adding # Pricing and Usage / # 定价与用量 H1 headings to the pricing pages is a good structural improvement that was previously missing and aids navigation and SEO. |
| Naming Standardisation | The shift from hyphenated to space-separated model names (MiniMax, Kimi, GPT variants) improves readability and aligns with most vendors' official display names. |
| Locale Parity | English and Chinese pricing tables are updated in tandem with matching values, maintaining full localisation parity for numerical data. |
Checklist Results
| Category | Items Checked | Pass | Fail | N/A | Notes |
|---|---|---|---|---|---|
| Correctness | 3 | 2 | 1 | 5 | Stray 白 character in 4 zh-Hans files (MJ-01) |
| Security | 8 | 8 | 0 | 0 | No code changes; no security surface exposed |
| Performance | 7 | 7 | 0 | 0 | Documentation only — not applicable |
| Code Quality | 6 | 5 | 1 | 0 | Inconsistent model name normalisation (MN-01) |
| Testing | 5 | 5 | 0 | 0 | Documentation only — no logic to test |
| Documentation | 6 | 4 | 2 | 0 | Missing DeepSeek V3.2 model page (MN-02); GPT-5.4 Pro restriction undocumented (MN-03) |
| Compatibility | 3 | 3 | 0 | 0 | No breaking changes to site structure |
| Observability | 2 | 2 | 0 | 0 | Not applicable to documentation |
Disclaimer
This is an automated code review. It supplements but does not replace human review. The reviewer analysed only the diff between main and ai-bankofai-patch-1. Runtime behaviour, integration testing, and deployment impact are not covered.
Report generated by Code Review Skill v1.0.0
Date: 2026-04-21
3 participants
This PR syncs Claude model pricing docs in
BofAI/docswith the latest updates fromBAI-labs/BAI-docs.Updated pages:
Synced pricing:
pricing-and-usageandclaude-opus-4-6had already been synced previously. This PR fills in the remaining model pages.