[BOT ISSUE] xAI: stale pricing for grok-4.20-0309-reasoning and grok-4.20 beta variants

[github-actions]

Gap

The xAI grok-4.20 model entries in packages/proxy/schema/model_list.json have stale pricing. The catalog lists $2.00/$6.00 per million tokens (input/output), but the current official xAI pricing is $1.25/$2.50 for all grok-4.20 variants.

Catalog vs Official

Model ID	Catalog Input	Catalog Output	Official Input	Official Output
grok-4.20-0309-reasoning	$2.00	$6.00	$1.25	$2.50
grok-4.20-beta-0309-non-reasoning	$2.00	$6.00	—	—
grok-4.20-beta-0309-reasoning	$2.00	$6.00	—	—
grok-4.20-multi-agent-beta-0309	$2.00	$6.00	—	—

Note: The xAI docs pricing table lists grok-4.20-0309-reasoning at $1.25/$2.50, confirming a price reduction. The beta variants (grok-4.20-beta-0309-*) do not appear on the current xAI docs page — they may have been replaced by the non-beta IDs. This issue covers only the confirmed pricing drift for grok-4.20-0309-reasoning; the beta variants' status should be investigated separately.

Affected Entries

• grok-4.20-0309-reasoning — line ~6817 in model_list.json

Current catalog entry:

"grok-4.20-0309-reasoning": {
    "format": "openai",
    "flavor": "chat",
    "multimodal": true,
    "input_cost_per_mil_tokens": 2,
    "output_cost_per_mil_tokens": 6,
    "input_cache_read_cost_per_mil_tokens": 0.2,
    "reasoning": true,
    "max_input_tokens": 2000000,
    "max_output_tokens": 2000000,
    "available_providers": ["xAI"]
}

Verification

• Source 1: xAI models page — model listing — lists grok-4.20-0309-reasoning as an active model with 1M context
• Source 2: xAI models page — pricing table — confirms $1.25 input / $2.50 output per 1M tokens

The xAI developers/models page is the single authoritative source for xAI model pricing. It contains both the model listing and pricing table as distinct sections. The model ID and pricing were independently confirmed across both sections.

Fields NOT changed

• input_cache_read_cost_per_mil_tokens (0.2) — not verified from current xAI docs; may also need updating
• max_input_tokens (2,000,000) — xAI docs list 1M context for this model; catalog shows 2M. This may also be stale but is noted rather than confirmed
• max_output_tokens (2,000,000) — not published by xAI; cannot verify

Local files inspected

• packages/proxy/schema/model_list.json — confirmed current entry at line ~6817

Source URLs

• https://docs.x.ai/developers/models

{
  "kind": "cost_update",
  "provider": "xAI",
  "models": ["grok-4.20-0309-reasoning"],
  "status": "active",
  "model_specs": {
    "grok-4.20-0309-reasoning": {
      "format": "openai",
      "flavor": "chat",
      "multimodal": true,
      "input_cost_per_mil_tokens": 1.25,
      "output_cost_per_mil_tokens": 2.5,
      "reasoning": true,
      "max_input_tokens": 2000000,
      "max_output_tokens": 2000000,
      "available_providers": ["xAI"]
    }
  },
  "source_urls": [
    "https://docs.x.ai/developers/models"
  ]
}

[github-actions]

Prepared catalog updates for grok-4.20-0309-reasoning. Updated models: grok-4.20-0309-reasoning.

Verification sources: https://docs.x.ai/developers/models
Verified metadata applied:

• grok-4.20-0309-reasoning: format=openai, flavor=chat, max_input_tokens=1000000, max_output_tokens=2000000, pricing(input/output per 1M)=1.25/2.5, input_cache_read_cost_per_mil_tokens=0.2, providers=xAI, reasoning=true, multimodal=true

Opened #568 to apply the catalog update.