Skip to content

feat(kimi): register k3-256k in the picker and the price catalog - #5447

Open
yuanyuanlove wants to merge 2 commits into
lidge-jun:devfrom
yuanyuanlove:feature-20260921-kimi-k3-256k
Open

feat(kimi): register k3-256k in the picker and the price catalog#5447
yuanyuanlove wants to merge 2 commits into
lidge-jun:devfrom
yuanyuanlove:feature-20260921-kimi-k3-256k

Conversation

@yuanyuanlove

@yuanyuanlove yuanyuanlove commented Sep 21, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

问题

Kimi 订阅端 /coding/v1/models 现在返回 k3-256k（与 kimi-for-coding[-highspeed]、k3 并列），但 opencodex 的 picker 和成本追踪目录都只有 k3 和 k3[1m]——用户选 k3-256k 后 usage 显示"无法估算"。

证据（2026-09-21 实测，非文档转述）

  • k3-256k 是服务端真实存在的 ID：POST /coding/v1/chat/completions 200。
  • 与 bare k3 是同一模型：同样长输入两边 prompt 都计 988 token（服务端脚手架一致），自我认知回答一致；K3 官方文档的计费表按 k3 一档计价。
  • k3[1m] 是 opencodex 本地别名（wire 上 strip 成 k3），而 k3-256k 是服务端原生 ID，不能靠 bracket-strip 合并。

改动

  • model-seeds.ts：KIMI_CODING_K3_MODELS 加 k3-256k，picker / 上下文窗口（262_144）/ 推理档位 / 锁定参数列表自动派生。
  • expected-prices.ts：kimi / kimi-code 各加 k3-256k 价格记录，同 KIMI_K3 档（input 3 / output 15 / cacheRead 0.3），source 注明 verified-derived 及实测依据。
  • parity 测试和 overlay membership 计数（126 → 128）同步更新。

验证

  • bun x tsc --noEmit 干净。
  • bun test tests/providers/provider-registry-parity.test.ts tests/codex-integration/codex-catalog.test.ts tests/usage/usage-cost.test.ts tests/providers/model-rename-migration.test.ts：539 pass / 0 fail。

Review readiness checklist

This PR stays in draft until every box below is ticked. Tick all four boxes once the requirements are met:

  • All CI tests are green on my local testing.

  • I pushed my PR to the latest dev commit.

  • I resolved all correct Codex and CodeRabbit findings.

  • My PR is ready for review.

Summary by CodeRabbit

  • New Features

    • Added support for the Kimi Coding K3 model’s 256K context variant, k3-256k.
    • Added pricing information for the new model variant across supported Kimi providers.
    • The new variant preserves the existing K3 model’s capabilities and context handling.

  • Tests

    • Updated model compatibility and pricing coverage to include the additional K3 variant and its provider-specific pricing entries.

@yuanyuan2026
feat(kimi): register k3-256k in the picker and the price catalog
5910394 
Live verification 260921: k3-256k is the same K3 served under the
explicit ceiling id - the same 988-token scaffold and identity answer as
bare k3 on the same input. The subscription endpoint lists it alongside
kimi-for-coding[-highspeed] and k3, but the opencodex picker and the
expected-prices overlay only knew k3 and k3[1m], so usage logged under
k3-256k showed as unestimable.

- KIMI_CODING_K3_MODELS gains k3-256k, so the picker, context windows
  (262_144, the advertised ceiling), reasoning ladder and locked-parameter
  lists all derive it automatically.
- expected-prices gains kimi/kimi-code entries at the same KIMI_K3 rate
  (input 3 / output 15 / cacheRead 0.3), sourced as verified-derived with
  the live probe note.
- Parity and overlay-membership tests updated for the new id.

Verified: provider-registry-parity + codex-catalog + usage-cost +
model-rename-migration 539 pass, tsc clean.
@github-actions

github-actions Bot commented Sep 21, 2026

Copy link
Copy Markdown
Contributor

✅ Deterministic PR hygiene checks passed.

@github-actions github-actions Bot added enhancement New feature or request review-ready labels Sep 21, 2026
@github-actions

github-actions Bot commented Sep 21, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

✅ READY

  • all PR quality gates passed; the review readiness checklist is complete.

Review readiness checklist

  • ✅ All CI tests are green on my local testing.
  • ✅ I pushed my PR to the latest dev commit.
  • ✅ I resolved all correct Codex and CodeRabbit findings.
  • ✅ My PR is ready for review.

✅ 4/4 boxes ticked.

This pull request is already Ready for Review.
The review-ready label marks this PR as ready; review automation runs independently.
Maintainers notified: @lidge-jun @Ingwannu

@coderabbitai

coderabbitai Bot commented Sep 21, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Review Change StackReview Change Stack

Understand this PR’s impact

Explore downstream dependencies and potential security impact with Blast Radius.

View blast radius →

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: Repository: lidge-jun/opencodex/.coderabbit.yaml

Review profile: ASSERTIVE

Plan: Advanced

Run ID: 779c4864-f60f-41ff-9785-8a254481bace

📥 Commits

Reviewing files that changed from the base of the PR and between 086a0f5 and c0484f4.

📒 Files selected for processing (1)
  • tests/usage/usage-cost.test.ts

Included review availability: Your plan provides up to 10 included reviews per hour; 8 remain after this review.

📝 Walkthrough

Walkthrough

The Kimi Coding registry now includes k3-256k. Both Kimi providers have matching expected-price overlays. Provider parity and overlay membership tests cover the added entries.

Changes

K3 256K Alias Support

Layer / File(s) Summary
Register the K3 256K alias
src/providers/registry/model-seeds.ts, tests/providers/provider-registry-parity.test.ts		 KIMI_CODING_K3_MODELS now includes k3-256k with a 262,144-token context window. The parity test includes the alias.
Add pricing overlays
src/usage/expected-prices.ts, tests/usage/usage-cost.test.ts		 The kimi and kimi-code providers receive verified-derived k3-256k overlays using K3 pricing. The membership test expects 128 overlays and checks both new keys.

Priority: ⬇️ Low

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Feature

🚥 Pre-merge checks | ✅ 5
✅ Passed checks (5 passed)
Check name Status Explanation
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 4…
Linked Issues check ✅ Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check ✅ Passed Check skipped because no linked issues were found for this pull request.
Description Check ✅ Passed Check skipped - CodeRabbit’s high-level summary is enabled.
Title check ✅ Passed The title clearly and concisely describes the primary changes: registering the k3-256k Kimi model in the picker and adding it to the price catalog. It is specific, relevant, and suitable for pull re…
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create a new PR

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@github-actions
github-actions Bot marked this pull request as draft September 21, 2026 10:09
@lidge-jun

lidge-jun commented Sep 21, 2026

Copy link
Copy Markdown
Owner

리뷰 · 우선순위 18 / 80

이 PR은 Kimi 구독 서버가 쓰는 모델 이름 k3-256k를, 고르는 목록과 비용표에 넣습니다.

지금까지 그 목록과 비용표에는 k3와 k3[1m]만 있었습니다. 사용자가 k3-256k를 고르면 사용량 화면이 가격을 못 정해서 "얼마인지 모르겠다"가 나왔습니다. 작성자는 2026-09-21에 직접 호출해 봤다고 본문에 적었습니다. k3-256k는 서버에 실제로 있고, 같은 글을 넣으면 k3와 같이 988토큰으로 세며, 답도 같았습니다. 공식 가격표도 k3 한 칸입니다.

씨앗 목록 KIMI_CODING_K3_MODELS에 k3-256k를 넣으면, 고르는 목록과 생각 단계와 잠긴 설정이 같이 따라갑니다. 창 크기는 262144입니다. 100만 토큰은 k3[1m]과 kimi-for-coding만 받고, k3-256k는 그 조건에 안 들어갑니다. 서버로 보낼 때 이름은 그대로 갑니다. 떼는 것은 끝의 대괄호 [1m]뿐이고, -256k는 안 뗍니다. 가격은 kimi와 kimi-code 둘 다 k3와 같습니다. 입력 3, 출력 15, 캐시 읽기 0.3입니다.

라인 tests/usage/usage-cost.test.ts:381 - 가격표 테스트는 줄 수를 126에서 128로만 올렸습니다. 꼭 있어야 하는 이름에는 kimi/k3와 kimi/k3[1m]는 있고, 새로 넣은 kimi/k3-256k와 kimi-code/k3-256k는 없습니다. 새 두 줄을 지우고 엉뚱한 두 줄을 넣어도, 줄 수가 128이면 이 테스트는 통과합니다.

메인테이너의 판단이 필요한 지점

고르는 목록에 k3와 k3-256k가 같이 나옵니다. 둘은 같은 K3이고 창도 256K입니다. 그래도 서버가 부르는 이름이 다르고, 비용은 그 이름을 그대로 찾습니다. 한 줄로 합치면 k3-256k 사용량이 다시 가격을 못 정합니다.

같은 실측 메모에는 kimi-for-coding-highspeed도 서버 목록에 있다고 적혀 있습니다. 그 이름은 이번 PR이 넣지 않습니다. 예전부터 빠져 있던 이름이라, 이번에 같이 넣을지는 따로 보면 됩니다.

너의 추천

그대로 병합해도 됩니다. 베이스는 dev이고, types.ts/config.ts 분할과 겹치지 않아서 닫을 중복 PR은 아닙니다. 병합 전에 가격 테스트의 이름 목록에 kimi/k3-256k와 kimi-code/k3-256k만 넣으면, 나중에 그 두 줄이 빠져도 테스트가 잡습니다.

이 댓글은 grok-bot이 작성했습니다

@yuanyuanlove
yuanyuanlove force-pushed the feature-20260921-kimi-k3-256k branch from 086a0f5 to 5910394 Compare September 21, 2026 10:16
@yuanyuanlove yuanyuanlove mentioned this pull request Sep 21, 2026
Closed
4 tasks
@yuanyuan2026
test(usage): pin k3-256k entries in the overlay membership list
c0484f4 
Per maintainer review on lidge-jun#5447: the overlay test only bumped the row
count 126->128, so dropping the new k3-256k rows and adding any other
two rows would still pass. Name kimi/k3-256k and kimi-code/k3-256k in
the mandatory list so the test fails if either row disappears.
@yuanyuanlove
yuanyuanlove marked this pull request as ready for review September 21, 2026 10:23
@github-actions
github-actions Bot marked this pull request as draft September 21, 2026 10:23
@yuanyuanlove
yuanyuanlove marked this pull request as ready for review September 21, 2026 10:25
@github-actions
github-actions Bot marked this pull request as draft September 21, 2026 10:25
@yuanyuanlove
yuanyuanlove marked this pull request as ready for review September 21, 2026 10:28
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@lidge-jun lidge-jun Awaiting requested review from lidge-jun lidge-jun is a code owner
@Ingwannu Ingwannu Awaiting requested review from Ingwannu Ingwannu is a code owner

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

enhancement New feature or request review-ready

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@yuanyuanlove @lidge-jun @yuanyuan2026