chore(pricing): refresh models.dev snapshot (adds gpt-5.6-sol; NEEDS REVIEW)

[miyaontherelay]

Summary

Refreshes crates/relayburn-sdk/data/models.dev.json via node scripts/update-pricing.mjs
(fetches https://models.dev/api.json).

Motivation: gpt-5.6-sol was absent from the snapshot, so real Codex turns were costed at
exactly $0.00 and rolled into the reported total. Observed:

model         turns  input      output  cost
gpt-5.6-sol   4      1,095,319  93,397  $0.00
warning: 4 turn(s) had no pricing for model(s): gpt-5.6-sol

The refresh adds it.

⚠️ This PR is deliberately red — please review before merging

Upstream also changed existing entries. Three tests fail, and they have been left failing
on purpose rather than updated silently, because two of them encode real cost semantics:

Test	Change
analyze::cost::tests::bills_reasoning_at_output_rate_for_claude_same_as_output (cost.rs:313)	claude-sonnet-4-6 reasoning_mode: SameAsOutput → Separate
analyze::pricing::tests::builtin_snapshot_has_gpt_5_5_pricing (pricing.rs:199)	gpt-5.5 input: 5.0 → 1.5
analyze::pricing::tests::builtin_snapshot_parses_and_has_anthropic_models (pricing.rs:178)	same gpt-5.5 input change

The reasoning-mode change is the one that matters. It asserts Claude bills reasoning tokens on a
separate tariff rather than at the output rate, which changes every computed Claude cost. That
is a product judgement about whether upstream is correct, not a test to be mechanically updated.

Decide, then update the assertions accordingly:

1. If upstream is right → update all three assertions and merge.
2. If upstream is wrong or unverified → pin/override the affected entries and merge only the
   gpt-5.6-sol addition.

Note on test design

These assertions hard-code live third-party prices, so every routine pricing:update breaks
CI. Worth a follow-up: assert structure and relationships (entry exists, rates > 0,
reasoning == output when mode is SameAsOutput) and cover exact values with a small pinned
fixture rather than the live snapshot.

Validation

• node scripts/update-pricing.mjs — succeeded; diff is 1 line (the snapshot is single-line JSON)
• cargo test --workspace — 831 passed, 3 failed (the three above)
• Baseline main — fully green, so this PR causes them

Split out of #498, which now contains only the unpriced-cost reporting fix and is green.

🤖 Generated with Claude Code

[coderabbitai]

Warning

Review limit reached

@miyaontherelay, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 57 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 63dab6aa-0d68-42f4-a172-c5c49aaf8249

📥 Commits

Reviewing files that changed from the base of the PR and between bafd6f1 and 2fafa38.

📒 Files selected for processing (1)

• CHANGELOG.md

📝 Walkthrough

Walkthrough

Updates the Unreleased changelog with the vendored models.dev snapshot refresh that adds gpt-5.6-sol pricing and corrects its previous $0.00 reporting.

Changes

Pricing documentation

Layer / File(s)	Summary
Record corrected model pricing CHANGELOG.md	Adds an Unreleased changelog bullet documenting the gpt-5.6-sol pricing update.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Possibly related PRs

• AgentWorkforce/burn#498: Updates the changelog for the same vendored pricing snapshot and unpriced model reporting correction.

Suggested reviewers: willwashburn

Poem

A rabbit found a price in the hay,
Where zero once had led astray.
gpt-5.6-sol now has its due,
In the changelog, fresh and true.
Hop, hop—pricing shines anew!

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the pricing snapshot refresh and the added gpt-5.6-sol entry.
Description check	✅ Passed	The description is directly about refreshing the pricing snapshot and explains the related test-impacting changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
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.

✨ Finishing Touches 💡 1

🛠️ Fix failing CI checks 💡

• Fix failing CI checks

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch chore/pricing-snapshot-refresh

---

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

• X
• Mastodon
• Reddit
• LinkedIn

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

[miyaontherelay]

@coderabbitai review

[coderabbitai]

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@CHANGELOG.md`:
- Line 7: Update the Unreleased pricing changelog entry to document all
user-visible changes from the vendored models.dev snapshot: the newly priced
gpt-5.6-sol model, the changed claude-sonnet-4-6 reasoning pricing, and the
changed gpt-5.5 input pricing.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 66f50dcc-8749-4237-8597-c49de0a7389a

📥 Commits

Reviewing files that changed from the base of the PR and between 962b2b7 and bafd6f1.

📒 Files selected for processing (2)

• CHANGELOG.md
• crates/relayburn-sdk/data/models.dev.json

[miyaontherelay]

Good catch — fixed in 2fafa38.

The entry named only the gpt-5.6-sol addition, so a reader would have learned that a new model is now priced without learning that costs for models they are already using changed in the same refresh. For a cost-attribution tool that is the more consequential half.

Updated:

- Pricing: refreshed the vendored models.dev snapshot. Adds `gpt-5.6-sol` (previously unpriced
  and reported as $0.00). Also changes pricing for models already in use: `claude-sonnet-4-6`
  reasoning switches from being billed at the output rate to a separate tariff
  (`reasoning_mode` `SameAsOutput` -> `Separate`), and `gpt-5.5` input drops from 5.0 to 1.5.
  Both alter previously-reported costs for existing usage.

Reviewer note: this PR is intentionally failing

Three tests are red and were left red rather than updated:

Test	Change
analyze::cost::tests::bills_reasoning_at_output_rate_for_claude_same_as_output	claude-sonnet-4-6 reasoning_mode SameAsOutput → Separate
analyze::pricing::tests::builtin_snapshot_has_gpt_5_5_pricing	gpt-5.5 input 5.0 → 1.5
analyze::pricing::tests::builtin_snapshot_parses_and_has_anthropic_models	same gpt-5.5 change

The reasoning-mode change is the blocker. It asserts Claude bills reasoning tokens on a separate tariff rather than at the output rate, which changes every computed Claude cost this tool reports. That is a judgement about whether upstream models.dev is correct, not a test to mechanically update — so it needs a human decision before merge.

Two ways to close this out:

1. Upstream is right → update all three assertions and merge.
2. Upstream is wrong or unverified → pin/override the affected entries and merge only the gpt-5.6-sol addition.

Related test-design issue

These assertions hard-code live third-party prices, so every routine pricing:update breaks CI. Worth a follow-up: assert structure and relationships (entry exists, rates > 0, reasoning == output when mode is SameAsOutput) and cover exact values with a small pinned fixture rather than the live snapshot. Otherwise this PR's situation recurs on every refresh.

Split from #498, which carries the unpriced-cost reporting fix and is green.