[backryun]

Summary

• add grok-4.6 with its official Responses-first capabilities, 500K context window, reasoning levels, vision/tool support, and pricing
• register grok-4.6 directly in the Grok Build (grok-cli) static fallback catalog and route it through the existing authenticated Responses transport
• refresh the official DeepSeek V4 Pro/Flash catalog, limits, and current pricing
• switch DeepSeek's default upstream transport from Chat Completions to native Responses
• expose DeepSeek's official Anthropic-compatible transport as a per-connection alternate
• preserve compatibility for older clients by reusing OmniRoute's existing Chat/Responses/Anthropic translators
• correct the existing xAI OAuth Grok 4.5 catalog entry to use its Responses endpoint

Why

The provider registry still described DeepSeek as Chat-Completions-only even though DeepSeek V4 now supports both the Responses API and an Anthropic-compatible Messages API. xAI also documents Grok 4.6 as a Responses-first model.

This keeps client compatibility at OmniRoute's protocol boundary: clients may continue sending Chat Completions or Responses requests while the router translates to the selected native upstream protocol. Grok Build receives an explicit static entry rather than inheriting the API-key catalog, while authenticated live model discovery remains enabled.

Validation

Passed:

• focused provider/transport tests: 47/47
• npm run typecheck:core
• npm run check:provider-consistency
• ESLint and Prettier on all changed files
• npm run check:doc-links
• git diff --check

Repository baseline failures observed outside this change:

• npm run test:vitest: existing modelSelectModalHelpers.ts parse error, AutoCombo weight drift, and missing NextIntlClientProvider test wrapper
• npm run lint: the same existing parser issue plus unrelated frozen/import and test typing failures
• pre-commit docs sync: package version 3.8.50 vs latest changelog release 3.8.49

Sources

• https://docs.x.ai/developers/models/grok-4-6
• https://docs.x.ai/developers/model-capabilities/text/comparison
• https://api-docs.deepseek.com/quick_start/pricing/
• https://api-docs.deepseek.com/guides/responses_api/
• https://api-docs.deepseek.com/guides/anthropic_api/

[diegosouzapw]

Verified against the PR's own sources — the data is accurate:

• DeepSeek pricing matches api-docs.deepseek.com/quick_start/pricing exactly as of 2026-08-13: flash $0.14 in / $0.0028 cached / $0.28 out, pro $0.435 in / $0.003625 cached / $0.87 out, 1M context / 384K max output.
• Grok 4.6 matches docs.x.ai pricing ($2.00 in / $0.50 cached / $6.00 out for <200K, 500K context).

One coordination note: this PR and #10201 have a modify/delete conflict on open-sse/config/providers/registry/xai-oauth/index.ts — #10201 deletes that module (co-locating OAuth into registry/xai/index.ts). The resolution is folding this PR's targetFormat: "openai-responses" line into the co-located block. Suggest landing #10201 first, then rebasing this one. Also please call out the DeepSeek default-transport switch in the description since it's a behavior change.

[backryun]

The PR description now explicitly calls out the DeepSeek default upstream transport switch in both the Summary and Why sections. I have also noted the merge order with #10201; if #10201 lands first, this branch will be rebased while preserving the openai-responses targetFormat in the co-located xAI OAuth block. Please take another look when convenient.

[diegosouzapw]

Validated in local merge-train .claude/worktrees/merge-train-20260816-000002-suite.log on devbox-vm-06-dev002 @ 8be62df65f61cc4cf29cdc3e6bc38173ffafa4e3 (FAST gates green: static + changed tests + vitest; daily full-suite run still required). Merged under the batch merge-train protocol (merge-gates §7). ⚠️ base-red inherited: #10523 (ServiceSupervisor order-dependent test, not touched by this batch).