[BOT ISSUE] OpenAI: update stale max_input_tokens for gpt-5, gpt-5-mini, and gpt-5-nano

[github-actions]

Summary

The max_input_tokens for gpt-5, gpt-5-mini, and gpt-5-nano is set to 272,000 in model_list.json, but OpenAI's official documentation states a 400,000-token context window for all three models. The catalog convention is max_input_tokens = context_window — confirmed by every other OpenAI model in the catalog:

Model	max_input_tokens	max_output_tokens	Convention
gpt-5.5	1,050,000	128,000	✅ max_input = context_window
gpt-5.4-mini	400,000	128,000	✅ max_input = context_window
gpt-5.4-nano	400,000	128,000	✅ max_input = context_window
gpt-5	272,000	128,000	❌ = context_window − max_output
gpt-5-mini	272,000	128,000	❌ = context_window − max_output
gpt-5-nano	272,000	128,000	❌ = context_window − max_output

The 272,000 value appears to have been incorrectly computed as 400,000 − 128,000 (context − max_output). Anthropic models also use the full context window (e.g., claude-sonnet-5 has max_input_tokens: 1000000 with max_output_tokens: 128000).

Affected Models

Model ID	Current max_input_tokens	Correct max_input_tokens	Line
gpt-5	272,000	400,000	~537
gpt-5-mini	272,000	400,000	~587
gpt-5-nano	272,000	400,000	~603

All other fields (format, flavor, pricing, max_output_tokens, providers) are correct and unchanged.

Also affected but not included in this issue (4th model exceeds the 3-model limit):

• gpt-5-pro: max_input_tokens 128,000 → should be 400,000 (same root cause; 128K = 400K − 272K max_output)

Verification Checklist

Check	Status	Detail
Cross-source	✅	Context window confirmed on individual model pages (gpt-5, gpt-5-mini, gpt-5-nano) — all state "400,000 context window"
Already fixed	✅	All three entries currently show max_input_tokens: 272000 in model_list.json
ID format	✅	Existing entries — no ID change
Duplicate check	✅	No open issue covers GPT-5 context window / max_input_tokens

Verification Notes

• "400,000 context window" is quoted directly from each model's official OpenAI docs page
• "128,000 max output tokens" is also quoted from the same pages — this field is already correct in the catalog
• The catalog convention (max_input_tokens = context_window) is consistent across all other major model families (GPT-5.4, GPT-5.5, Claude, Gemini)

Local Files Inspected

• packages/proxy/schema/model_list.json — confirmed current values at lines ~537, ~587, ~603
• packages/proxy/schema/models.ts — max_input_tokens described as "The model supports a maximum input token limit"
• packages/proxy/scripts/sync_models.ts — none of these models are in SYNC_PRESERVED_FIELDS

{
  "kind": "stale_metadata",
  "provider": "openai",
  "models": ["gpt-5", "gpt-5-mini", "gpt-5-nano"],
  "status": "active",
  "model_specs": {
    "gpt-5": {
      "format": "openai",
      "flavor": "chat",
      "multimodal": true,
      "input_cost_per_mil_tokens": 1.25,
      "output_cost_per_mil_tokens": 10,
      "input_cache_read_cost_per_mil_tokens": 0.125,
      "displayName": "GPT-5",
      "reasoning": true,
      "max_input_tokens": 400000,
      "max_output_tokens": 128000,
      "available_providers": ["openai", "azure"]
    },
    "gpt-5-mini": {
      "format": "openai",
      "flavor": "chat",
      "multimodal": true,
      "input_cost_per_mil_tokens": 0.25,
      "output_cost_per_mil_tokens": 2,
      "input_cache_read_cost_per_mil_tokens": 0.025,
      "displayName": "GPT-5 mini",
      "reasoning": true,
      "max_input_tokens": 400000,
      "max_output_tokens": 128000,
      "available_providers": ["openai", "azure"]
    },
    "gpt-5-nano": {
      "format": "openai",
      "flavor": "chat",
      "multimodal": true,
      "input_cost_per_mil_tokens": 0.05,
      "output_cost_per_mil_tokens": 0.4,
      "input_cache_read_cost_per_mil_tokens": 0.005,
      "displayName": "GPT-5 nano",
      "reasoning": true,
      "max_input_tokens": 400000,
      "max_output_tokens": 128000,
      "available_providers": ["openai", "azure"]
    }
  },
  "source_urls": [
    "https://developers.openai.com/api/docs/models/gpt-5",
    "https://developers.openai.com/api/docs/models/gpt-5-mini",
    "https://developers.openai.com/api/docs/models/gpt-5-nano"
  ]
}

[github-actions]

Prepared catalog updates for gpt-5, gpt-5-mini, gpt-5-nano. Updated models: gpt-5, gpt-5-mini, gpt-5-nano.

Verification sources: https://developers.openai.com/api/docs/models/gpt-5, https://developers.openai.com/api/docs/models/gpt-5-mini, https://developers.openai.com/api/docs/models/gpt-5-nano
Verified metadata applied:

• gpt-5: format=openai, flavor=chat, displayName=GPT-5, max_input_tokens=400000, max_output_tokens=128000, pricing(input/output per 1M)=1.25/10, input_cache_read_cost_per_mil_tokens=0.125, providers=openai|azure, reasoning=true, multimodal=true
• gpt-5-mini: format=openai, flavor=chat, displayName=GPT-5 mini, max_input_tokens=400000, max_output_tokens=128000, pricing(input/output per 1M)=0.25/2, input_cache_read_cost_per_mil_tokens=0.025, providers=openai|azure, reasoning=true, multimodal=true
• gpt-5-nano: format=openai, flavor=chat, displayName=GPT-5 nano, max_input_tokens=400000, max_output_tokens=128000, pricing(input/output per 1M)=0.05/0.4, input_cache_read_cost_per_mil_tokens=0.005, providers=openai|azure, reasoning=true, multimodal=true

Included in the daily batch PR: #917