fix: update xAI model metadata for grok-4.20-0309-reasoning

[github-actions]

fix: update xAI model metadata for grok-4.20-0309-reasoning

Closes #567

Source issue: #567

Summary

Field	Value
Provider	xai
Primary model	grok-4.20-0309-reasoning
Changed models	grok-4.20-0309-reasoning
Added models	None
Updated models	grok-4.20-0309-reasoning
Verification sources	1

Verified metadata

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
grok-4.20-0309-reasoning			xAI	openai	chat	input=1000000, output=2000000	in/out=1.25/2.5 per 1M; cache read=0.2 per 1M	multimodal=true; reasoning=true

Verification notes

Verification

Official source verification

• Source: xAI developers/models page
  • Confirms grok-4.20-0309-reasoning as an active model
  • Confirms pricing: $1.25/1M input, $2.50/1M output
  • Confirms context window: 1M tokens
  • Does NOT publish max output tokens
  • Does NOT publish cache pricing (read or write)
  • Does NOT publish multimodal capability per-model (page mentions image input support generically)
  • Lists non-beta replacements: grok-4.20-0309-non-reasoning and grok-4.20-multi-agent-0309

Fields verified and changed

Field	Old Value	New Value	Source
input_cost_per_mil_tokens	2	1.25	https://docs.x.ai/developers/models
output_cost_per_mil_tokens	6	2.5	https://docs.x.ai/developers/models
max_input_tokens	2000000	1000000	https://docs.x.ai/developers/models (listed as "1M")

Fields NOT changed (not verifiable from official source)

• input_cache_read_cost_per_mil_tokens (0.2) — xAI docs do not publish cache pricing; retaining existing value
• max_output_tokens (2000000) — xAI docs do not publish max output tokens; retaining existing value
• multimodal (true) — xAI docs mention image input support generically but do not confirm per-model; retaining existing value
• reasoning (true) — implied by model name suffix; retaining existing value

sync_models (LiteLLM) cross-check

The LiteLLM model_prices_and_context_window_backup.json catalog contains no entries for any xAI or grok model. There are zero entries with litellm_provider containing "xai" and zero entries with "grok" in the key name. Therefore no numeric cross-check is possible. All proposed values are sourced directly from the official xAI documentation at https://docs.x.ai/developers/models.

Models NOT covered by this issue

The following beta variants from the original issue table are not updated because they no longer appear on the official xAI models page:

• grok-4.20-beta-0309-non-reasoning — replaced by grok-4.20-0309-non-reasoning on official docs
• grok-4.20-beta-0309-reasoning — no longer listed; grok-4.20-0309-reasoning (non-beta) is the current active model
• grok-4.20-multi-agent-beta-0309 — replaced by grok-4.20-multi-agent-0309 on official docs

These beta models may need deprecation or removal in a separate issue.

sync_models vs proposed update

sync_models cross-check found differences. Official provider verification was used for the applied values, and sync_models discrepancies are listed below for review.

Model	Field	Proposed update	sync_models	sync_models source models
grok-4.20-0309-reasoning	max_input_tokens	1000000	2000000	xai/grok-4.20-0309-reasoning
grok-4.20-0309-reasoning	input_cost_per_mil_tokens	1.25	2	xai/grok-4.20-0309-reasoning
grok-4.20-0309-reasoning	output_cost_per_mil_tokens	2.5	6	xai/grok-4.20-0309-reasoning

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
ai-proxy	Ready	Preview, Comment	May 19, 2026 6:16pm