feat(xai): cherry-pick grok-4.3 model entries onto main (#27154)

[mateo-berri]

Relevant issues

Cherry-picks #27154 from litellm_internal_staging onto main so that xai/grok-4.3 and xai/grok-4.3-latest show up in /v1/models for users running off main.

Linear ticket

Resolves LIT-2914

Pre-Submission checklist

• I have Added testing in the tests/test_litellm/ directory — tests/test_litellm/test_xai_grok_4_3_model_metadata.py (3 tests: pricing/capabilities for xai/grok-4.3 and xai/grok-4.3-latest, plus parity between the canonical and bundled-backup cost maps).
• My PR passes all unit tests on make test-unit (the new test file passes locally — see proof of fix below).
• My PR's scope is as isolated as possible — it only adds the two grok-4.3 model entries and the matching test.
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Screenshots / Proof of Fix

$ uv run pytest tests/test_litellm/test_xai_grok_4_3_model_metadata.py -v
============================= test session starts ==============================
collected 3 items

tests/test_litellm/test_xai_grok_4_3_model_metadata.py::test_xai_grok_4_3_backup_matches_main PASSED [ 33%]
tests/test_litellm/test_xai_grok_4_3_model_metadata.py::test_xai_grok_4_3_model_info[xai/grok-4.3-latest] PASSED [ 66%]
tests/test_litellm/test_xai_grok_4_3_model_metadata.py::test_xai_grok_4_3_model_info[xai/grok-4.3] PASSED [100%]

============================== 3 passed in 0.14s ===============================

Type

🆕 New Feature

Changes

Straight cherry-pick of fee5900 (PR #27154) — no conflicts, no edits. Adds xai/grok-4.3 and xai/grok-4.3-latest to both the canonical model_prices_and_context_window.json and the bundled litellm/model_prices_and_context_window_backup.json, and includes the metadata/parity test.

Pricing/specs sourced from xAI's published model metadata (https://docs.x.ai/docs/models):

• input: $1.25 / 1M tokens (≤200k), $2.50 / 1M tokens (>200k)
• output: $2.50 / 1M tokens (≤200k), $5.00 / 1M tokens (>200k)
• cached input: $0.20 / 1M tokens (≤200k), $0.40 / 1M tokens (>200k)
• context window: 1,000,000 tokens
• capabilities: vision, reasoning, function calling, structured outputs, prompt caching, web search

Slack Thread

  

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[codspeed-hq]

Merging this PR will not alter performance

✅ 16 untouched benchmarks

---

_{Comparing litellm_hotfix_cherry-pick-grok-4.3-d843 (4b587dc) with main (6ff668c)}

[greptile-apps]

Greptile Summary

Cherry-picks the xai/grok-4.3 and xai/grok-4.3-latest model entries from the internal staging branch onto main, adding them to both the canonical and bundled cost-map JSON files so they appear in /v1/models.

• Pricing and capabilities match xAI's published documentation: $1.25/$2.50 per 1M tokens (≤200k), tiered higher-context pricing above 200k, 1M-token context window, vision, reasoning, function calling, prompt caching, and web search.
• Both JSON files (model_prices_and_context_window.json and the bundled backup) are updated identically and verified by the new parity test.
• New test file (test_xai_grok_4_3_model_metadata.py) validates pricing fields, capability flags, and provider routing using only local file reads — no network calls.

Confidence Score: 4/5

Safe to merge — the change is purely additive (two new JSON entries and a new test file), with no modifications to existing entries or logic.

The added entries are accurate against xAI's published docs and the two JSON files are kept in sync. The one gap is that xAI lists a third alias (grok-latest) for this model that is absent from both cost maps, which would cause a model-not-found error for callers using that alias.

Both model_prices_and_context_window.json and its backup are missing the xai/grok-latest alias entry documented by xAI.

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Adds xai/grok-4.3 and xai/grok-4.3-latest entries with correct pricing and capabilities; missing the third published alias grok-latest
litellm/model_prices_and_context_window_backup.json	Backup file kept in sync with canonical file; same two entries added, same missing grok-latest alias
tests/test_litellm/test_xai_grok_4_3_model_metadata.py	New test file that reads JSON files locally and calls get_llm_provider; no network calls, CI-safe

_{Reviews (1): Last reviewed commit: "feat(xai): add grok-4.3 and grok-4.3-lat..." | Re-trigger Greptile}

[greptile-apps]

Missing grok-latest alias entry

xAI's published model page lists three aliases for this model: grok-4.3-latest, and grok-latest (alongside the canonical grok-4.3). This PR adds the first two but omits xai/grok-latest. Users who call xai/grok-latest (mirroring the provider alias) will get a model-not-found error. Consider adding the third alias entry with identical pricing/capabilities, or document the decision to skip it.