feat(xai): add grok-4.3 and grok-4.3-latest to model_prices_and_context_window.json

[ishaan-berri]

What

Adds the missing grok-4.3 flagship (and its -latest alias) to model_prices_and_context_window.json.

Why

xAI's docs/models page now leads with Grok 4.3 and explicitly says:

"We strongly recommend all API callers use grok-4.3. It is the most intelligent and fastest model we've built."

But it isn't currently listed in model_prices_and_context_window.json, so cost tracking, context-window limits, and capability flags don't work for users hitting it through litellm. The other 30+ Grok models are all there.

Pricing & specs (sourced from xAI's published model metadata)

Field	Value
input_cost_per_token	$1.25 / 1M (≤200k tokens)
input_cost_per_token_above_200k_tokens	$2.50 / 1M
output_cost_per_token	$2.50 / 1M (≤200k tokens)
output_cost_per_token_above_200k_tokens	$5.00 / 1M
cache_read_input_token_cost	$0.20 / 1M (≤200k)
cache_read_input_token_cost_above_200k_tokens	$0.40 / 1M
max_tokens / max_input_tokens / max_output_tokens	1,000,000
Capabilities	vision, reasoning, function calling, structured outputs, prompt caching, web search

Verified against known anchors before encoding: xAI publishes raw token prices as fixed-point integers where 30000 corresponds to grok-4's known $3/M and 3000 corresponds to grok-3-mini's $0.30/M — so the unit is value / 10000 per 1M tokens. grok-4.3's published values are 12500 / 25000 / 25000 / 50000 for prompt-short / completion-short / prompt-long / completion-long respectively, which decode to the table above.

The 200k-token long-context threshold matches xAI's longContextThreshold: 200000 for grok-4.3 and uses the existing *_above_200k_tokens keys already present in this file.

Diff

42-line surgical insertion — only adds two new top-level entries, doesn't touch any existing model. JSON validity verified locally.

Out of scope

While auditing, I also noticed a couple of items not included in this PR (happy to follow up):

• Missing aliases: xai/grok-4-fast, xai/grok-4-fast-non-reasoning-latest, xai/grok-4-fast-reasoning-latest, xai/grok-3-fast.
• xai/grok-4.20-* entries currently priced at $2 / $6 per 1M, but xAI's published metadata is $1.25 / $2.50 short-ctx and $2.50 / $5.00 long-ctx — could be a stale snapshot worth refreshing in a separate PR.

---

🤖 Opened by shin-watcher in response to "check if we are missing any xAI models, if we are add them as a PR". Source: https://docs.x.ai/docs/models.

[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.

---

shin-watcher seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{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 shin-watcher/add-xai-grok-4-3 (1c31e2c) with main (934ecdc)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[greptile-apps]

Greptile Summary

This PR makes a surgical, 42-line addition of xai/grok-4.3 and xai/grok-4.3-latest to the model prices/context-window registry. The JSON is structurally valid, pricing values are mathematically consistent with the PR description (e.g. 1.25e-06 = $1.25/M tokens), capability flags match the pattern of comparable recent xAI models (grok-4-1-fast), and the -latest alias is a full-copy entry consistent with how xai/grok-4-latest and other aliases are handled throughout the file.

Confidence Score: 4/5

Safe to merge — purely additive JSON data entry with no logic changes and no risk to existing models.

No P0 or P1 issues found. Pricing arithmetic checks out ($1.25e-6, $2.5e-6, etc.), the JSON structure mirrors existing xAI entries, and the -latest alias pattern (full-copy entry) is consistent with how other aliases like xai/grok-4-latest are handled. The only uncertainty is whether the source URL (https://docs.x.ai/docs/models/grok-4.3) is live and whether all capability flags (e.g. audio input) are correctly captured — both are P2-level and unverifiable without docs access.

No files require special attention beyond confirming the xAI model page is live and pricing is current.

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Adds two new top-level entries — xai/grok-4.3 and xai/grok-4.3-latest — with pricing, context-window limits, and capability flags; JSON structure and pricing values are consistent with the file's existing xAI entries and the PR author's sourced math.

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

[krrish-berri-2]

merge to internal staging, not main @ishaan-berri

[oss-pr-review-agent-shin]

🤖 litellm-agent: This PR's commits have been merged into the staging branch litellm_agent_oss_staging_05_05_2026.

Staging PR: #27162