Skip to content

chore(models): price the November 16 catalog additions - #220

Merged
pierreraby merged 1 commit into
patlux:mainfrom
sandexzx:chore/models-pricing-2026-11-16
Nov 17, 2026
Merged

pierreraby merged 1 commit into
patlux:mainfrom
sandexzx:chore/models-pricing-2026-11-16

Conversation

@sandexzx

Copy link
Copy Markdown
Contributor

Summary

/provider/v1/models advertised 100 models on 2026-11-16 (was 98 on 2026-11-13). Two of them had no reviewed price, so pi showed a zero estimate plus an unknown-price warning on every request that used them, and the daily catalog workflow would fail its pricing gate.

Reviewed against https://commandcode.ai/docs/resources/pricing-limits. The page column order is context, input, output, cache read, cache write; a — column means 0:

Model Context Input Output Cache read Cache write Endpoints
Qwen/Qwen4-Max 1M 1.50 4.50 0.075 1.875 /chat/completions, /responses
tencent/hy4-flash 1M 0.07 0.281 0.003 — /chat/completions, /responses

Neither row carries an "Available on …" plan note or a long-context tier, and the promotions table publishes no deal for them, so no TEMPORARY_PRICING entry is involved. The page ids are qwen-4-max and tencent/hy4-flash.

Sources and policy

  • Official pricing page: https://commandcode.ai/docs/resources/pricing-limits — the two rows above, single published rates, no plan note, no tier, no promotion.
  • Public Provider API: https://api.commandcode.ai/provider/v1/models — 100 models; the two additions are the only delta against the 2026-11-13 snapshot, and no model left the API.
  • Both live entries advertise only id, name, context_length and supported_endpoints (/chat/completions, /responses); contexts are 1,000,000 and 1,024,000 tokens.
  • CLI metadata is current at command-code@1.89.1 and does not list either model. Both therefore keep the conservative text-only, non-reasoning defaults with a 65,536-token output limit until upstream publishes capabilities.

Changes

Five files: two MODEL_COSTS entries plus the verification date in src/pricing.ts, regenerated API model-ID and official pricing fixtures (100 models, verified 2026-11-16), an explicit rate regression test for both models, and a CHANGELOG Unreleased entry. No runtime, routing, workflow, CLI-pin or other model rates changed.

Validation

  • Before: npm run check:commandcode-pricing exited 1 with missing-local-price for exactly Qwen/Qwen4-Max and tencent/hy4-flash.
  • After: live pricing check PASS, 100 models checked, no drift, snapshot verified 2026-11-16.
  • npm run check:commandcode-catalog → metadata Current at command-code@1.89.1 (78 image, 87 reasoning, 72 efforts, 10 output limits), zero diff.
  • npm run test:pricing 15/15 PASS (including both new rates and the provenance date); npm run test:models 32/32 PASS plus the metadata checker 19/19 PASS.
  • npm run typecheck, npm run format:check and git diff --check PASS.
  • Full npm test PASS against real hosts, including the host-contract and endpoint-routing scenarios for both new models.

Limits

  • No live account login or billable inference was performed. Display prices are reviewed from the public page; the Command Code Usage page remains authoritative per request.
  • Static capability metadata is unchanged at command-code@1.89.1, so both models remain text-only without reasoning or effort flags even though a future CLI release may advertise more. That is catalog work for a future refresh, not part of this price update.
  • The historical HTML pricing fixture is not regenerated, as in previous refreshes; no assertion reads it.

@pierreraby pierreraby left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed head eaf3cc8 against 3b60d88 in an isolated worktree. The two prices match the official pricing page, including cache column order, with no additional tier or promotion for these rows. Local typecheck, pricing/models/pricing-check/cost tests, entrypoint typecheck, formatting, and full npm test with both Pi and OMP required passed against mocks. A second fresh-context read-only review found no blocking defect in this pricing change.

Upstream moved after this snapshot: the public API now lists 101 models (Flint is new), and command-code@1.89.2 adds capabilities for Qwen4-Max, Hy4 Flash and Flint. The live pricing and catalog gates therefore still require a follow-up; I will handle that separately and update the Unreleased wording before release.

One validation clarification: the host-contract and routing suites passed, but they do not contain dedicated scenarios named for these two new model IDs.

I am currently handling day-to-day maintenance solo; keeping this merge limited to the reviewed price additions. No billable inference was performed.

@pierreraby
pierreraby merged commit 7d74f45 into patlux:main Nov 17, 2026
15 checks passed
pierreraby added a commit that referenced this pull request Nov 17, 2026
Sync command-code 1.89.2 capabilities and review the 101-model API snapshot. Add undated free-tier pricing for Flint and explicit model/routing and pricing-page ID regressions. Correct the Unreleased description following #220.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants