feat(models): add DeepSeek V4 Pro to the model picker

[zachdive]

Summary

• Add deepseek/deepseek-v4-pro-0813 to PARAMETRIC_MODELS — tools and reasoning confirmed via OpenRouter's endpoint metadata; the model is text-only, so supportsVision is false
• Price it at $0.435/M input, $0.87/M output, cache reads at $0.003625/M per OpenRouter; cacheWrite is set to the input rate explicitly since DeepSeek has no cache-write surcharge and the fallback would bill 1.25x
• Add the DeepSeek whale mark to ProviderLogo (official monochrome path from simple-icons), since DeepSeek is a new provider in the picker

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

Test plan

• npm run typecheck passes
• Verify DeepSeek V4 Pro appears in the model picker with its logo and completes a chat/CAD generation

🤖 Generated with Claude Code

---

Summary by cubic

Adds deepseek/deepseek-v4-pro-0813 to the model picker with correct pricing and a DeepSeek provider icon, enabling a low‑cost 1M‑token MoE model that was previously unavailable. The model is text‑only.

Details for review

• Add to PARAMETRIC_MODELS: tools/thinking enabled; vision disabled.
• Set MODEL_PRICES: input $0.435/M, output $0.87/M, cacheRead $0.003625/M; set cacheWrite to $0.435/M to avoid the default 1.25× surcharge.
• Add DeepSeek whale mark to ProviderLogo.
• No routing or env changes; OpenRouter handles non‑Anthropic/Google IDs.

^{Written for commit cddaf58. Summary will update on new commits.}

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
cadam	Ready	Preview	Aug 12, 2026 10:01pm

[supabase]

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 ↗︎.

[cubic-dev-ai]

No issues found across 3 files

_{Re-trigger cubic}

[greptile-apps]

Greptile Summary

Adds DeepSeek V4 Pro as a text-only parametric model routed through OpenRouter.

• Adds DeepSeek metadata and capability flags to the model picker.
• Adds the provider's monochrome logo.
• Adds explicit input, output, cache-read, and cache-write billing rates.

Confidence Score: 4/5

The PR appears safe to merge after addressing the non-blocking source-format violation in the new provider icon.

The model identifier, provider key, capability metadata, OpenRouter routing, and billing key align; the only accepted concern is the oversized SVG source line.

Files Needing Attention: src/components/ProviderLogo.tsx

Important Files Changed

Filename	Overview
src/components/ProviderLogo.tsx	Adds a matching DeepSeek provider mark, but its path violates the repository's maximum line-length convention.
src/lib/utils.ts	Adds the DeepSeek model with consistent provider metadata and text-only capability marking.
src/server/aiChat.ts	Adds a matching pricing entry with an explicit input-rate cache-write price to avoid the default surcharge.

_{Reviews (1): Last reviewed commit: "feat(models): add DeepSeek V4 Pro to the..." | Re-trigger Greptile}

[greptile-apps]

Oversized DeepSeek SVG line

The complete DeepSeek path is written on one source line exceeding the repository's 100-character limit, making the icon data unnecessarily difficult to inspect and edit.

Context Used: .cursor/rules/code-style.mdc (source)

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

[zachdive]

This matches the file's existing convention: the OpenAI and Anthropic paths above are also single ~1,500-character lines. SVG path data is a single string literal, so Prettier can't wrap it, and splitting it into concatenated chunks would make the icon harder to diff against its source SVG. Keeping icon paths on one line in this file is intentional; please treat it as an accepted exception to the 100-character rule.