Skip to content

feat(models): add Qwen3.8 Max to the model picker - #248

Open
zachdive wants to merge 1 commit into
masterfrom
bold-jackal
Open

feat(models): add Qwen3.8 Max to the model picker#248
zachdive wants to merge 1 commit into
masterfrom
bold-jackal

Conversation

@zachdive

@zachdive zachdive commented Aug 4, 2026

Copy link
Copy Markdown
Contributor

Summary

  • Add qwen/qwen3.8-max to PARAMETRIC_MODELS — tools, thinking, and vision all confirmed via OpenRouter's model metadata
  • Price it per OpenRouter's listing: $2/M input, $6/M output, cache reads at $0.25/M (12.5% of input), cache writes at $2.50/M (1.25x input)
  • Add the Qwen provider mark to ProviderLogo so picker rows render a logo like the other providers

No routing changes needed: non-Anthropic/Google ids already flow through OpenRouter.

Test plan

  • npm run typecheck passes
  • eslint clean on the three edited files
  • Verify Qwen3.8 Max appears in the model picker with its logo and generates a model end to end

🤖 Generated with Claude Code


Summary by cubic

Add qwen/qwen3.8-max to the model picker with correct pricing and a Qwen provider logo. Supports tools, thinking, and vision via OpenRouter.

  • New Features
    • Added to PARAMETRIC_MODELS with tools, thinking, and vision enabled.
    • Set MODEL_PRICES: $2/M input, $6/M output, cache read $0.25/M, cache write $2.50/M.
    • Added Qwen icon to ProviderLogo; routing unchanged (handled via OpenRouter).

Written for commit 6eb3a4d. Summary will update on new commits.

Review in cubic

Add qwen/qwen3.8-max to PARAMETRIC_MODELS (tools, thinking, and vision
all confirmed via OpenRouter's model metadata) and price it at $2/M
input, $6/M output, cache reads at $0.25/M (12.5% of input), cache
writes at $2.50/M (1.25x input). Also add the Qwen provider mark to
ProviderLogo so picker rows render a logo like the other providers.

No routing changes needed: non-Anthropic/Google ids already flow through
OpenRouter.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@vercel

vercel Bot commented Aug 4, 2026

Copy link
Copy Markdown

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

Project Deployment Actions Updated (UTC)
cadam Ready Ready Preview Aug 4, 2026 12:12am

Request Review

@supabase

supabase Bot commented Aug 4, 2026

Copy link
Copy Markdown

This pull request has been ignored for the connected project sgprnbvihmydyrzvkcir because there are no changes detected in supabase directory. You can change this behaviour in Project Integrations Settings ↗︎.


Preview Branches by Supabase.
Learn more about Supabase Branching ↗︎.

@greptile-apps

greptile-apps Bot commented Aug 4, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

Adds Qwen3.8 Max to the selectable parametric models.

  • Declares its tools, thinking, and vision capabilities.
  • Adds OpenRouter token and cache pricing.
  • Adds a Qwen provider icon for model-selection UI.

Confidence Score: 4/5

The malformed Qwen SVG path should be corrected before merging because the newly added provider mark can render clipped or distorted.

The model catalog and pricing additions are coordinated, but the icon's relative move commands put later subpaths outside ProviderLogo's fixed 24-by-24 viewBox.

Files Needing Attention: src/components/ProviderLogo.tsx

Important Files Changed

Filename Overview
src/components/ProviderLogo.tsx Adds the Qwen mark, but relative subpath moves place substantial geometry outside the fixed SVG viewBox.
src/lib/utils.ts Adds Qwen3.8 Max and its advertised capabilities to the parametric model catalog.
src/server/aiChat.ts Adds internally consistent input, output, cache-read, and cache-write pricing for the new model.

Reviews (1): Last reviewed commit: "feat(models): add Qwen3.8 Max to the mod..." | Re-trigger Greptile

),
'Moonshot AI': <path d="M21 12.79A9 9 0 1 1 11.21 3a7 7 0 0 0 9.79 9.79z" />,
Qwen: (
<path d="M23.919 14.545 20.817 9.17l1.47-2.544a.56.56 0 0 0 0-.566l-1.633-2.83a.57.57 0 0 0-.49-.283h-6.207L12.487.402a.57.57 0 0 0-.49-.284H8.732a.56.56 0 0 0-.49.284L5.139 5.775h-2.94a.56.56 0 0 0-.49.284L.077 8.887a.56.56 0 0 0 0 .567L3.18 14.83l-1.47 2.545a.56.56 0 0 0 0 .566l1.634 2.83a.57.57 0 0 0 .49.283h6.205l1.47 2.545a.57.57 0 0 0 .49.284h3.266a.57.57 0 0 0 .49-.284l3.104-5.375h2.94a.57.57 0 0 0 .49-.283l1.634-2.828a.55.55 0 0 0-.004-.568M8.733.686l1.634 2.828-1.634 2.828H21.8L20.164 9.17H7.425L5.63 6.06Zm1.306 19.801-6.205-.002 1.634-2.83h3.265L2.201 6.344h3.267q3.182 5.517 6.367 11.032zm10.124-5.66L18.53 12l-6.532 11.315-1.634-2.83c2.129-3.673 4.25-7.351 6.373-11.028h3.592l3.102 5.374z" />

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

P1 Qwen subpaths exceed viewBox

When the Qwen logo renders in the model picker or retry menu, its relative m commands place later subpaths outside the fixed 0 0 24 24 viewBox, causing the provider mark to render clipped or distorted.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

False positive — verified empirically. The path is verbatim from Simple Icons (which normalizes every icon to a 24x24 viewBox), and parsing every anchor/control point in the path gives a bounding box of x 0.077–23.923, y 0.118–23.883, fully inside 0 0 24 24. Rendering the SVG confirms the mark draws complete and unclipped.

The review's reasoning misses that z resets the current point to the start of the current subpath, so the relative m commands that follow it land where intended (x≈10.0 and x≈20.2), not outside the viewBox.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

You're right, and the technical explanation is sound. After z, the current point resets to the start of the subpath, so the relative m that follows anchors from that reset position — not from wherever the path was before z. Combined with Simple Icons' guaranteed 24×24 normalization and your empirical bounding-box check (0.077–23.923, 0.118–23.883), the path is fully within the viewBox. The original concern was a false positive.

@cubic-dev-ai cubic-dev-ai Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

No issues found across 3 files

Re-trigger cubic

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.

1 participant