[BOT ISSUE] Mistral: add missing cached input pricing for flagship models

[github-actions]

Gap

All current Mistral flagship models are missing input_cache_read_cost_per_mil_tokens in the catalog, despite Mistral publishing explicit cached input pricing on their official pricing page. The input and output costs are correct — only the cached input field is absent.

This affects cost tracking for any Mistral API user leveraging prompt caching, which Mistral supports across all flagship models.

Affected Models (top 3)

Model ID	Catalog Line	Current input_cache_read_cost_per_mil_tokens	Correct Value	Source
mistral-large-latest	11064	(missing)	$0.05	Mistral Pricing
mistral-medium-3.5	11091	(missing)	$0.15	Mistral Pricing
mistral-small-latest	11299	(missing)	$0.015	Mistral Pricing

Additional affected entries

The same gap applies to all dated variants and other current Mistral models:

Model ID	Correct input_cache_read_cost_per_mil_tokens
mistral-large-2512	$0.05
mistral-small-2603	$0.015
codestral-2508 / codestral-latest	$0.03
ministral-14b-latest / ministral-14b-2512	$0.02
ministral-8b-latest / ministral-8b-2512	$0.015
ministral-3b-latest / ministral-3b-2512	$0.01

Note: the recently added zai-glm-5-2 entry (issue #1097) correctly includes input_cache_read_cost_per_mil_tokens: 0.14, confirming this field is expected for Mistral models.

Recommended update (top 3)

"mistral-large-latest": {
  "input_cache_read_cost_per_mil_tokens": 0.05
}

"mistral-medium-3.5": {
  "input_cache_read_cost_per_mil_tokens": 0.15
}

"mistral-small-latest": {
  "input_cache_read_cost_per_mil_tokens": 0.015
}

Verification Checklist

Check	Status	Detail
Cross-source	✅	Cached pricing confirmed on Mistral pricing page (explicit "Cached Input" column) AND models confirmed active on Mistral models overview
Already-fixed check	✅	Verified on current main — all three entries lack input_cache_read_cost_per_mil_tokens
ID format check	✅	Existing catalog entries — no ID changes needed
Duplicate issue check	✅	No open or closed issue covers Mistral cached input pricing. Closed #1097 added zai-glm-5-2 (with cache pricing). Closed #1120 fixed zai-org/GLM-5.2 cache pricing. Neither covers Mistral's own models
Not embedding/vision-only	✅	All are generalist chat models

Verification Notes

Field	Source	Notes
Cached input prices	Mistral pricing	Explicit "Cached Input" column on pricing table. Pattern: cached = 10% of standard input across all models
Models active	Mistral models overview	All listed as current; none in deprecated table
Input/output costs correct	Catalog vs pricing page	Standard input and output costs already match between catalog and pricing page

Local Files Inspected

• packages/proxy/schema/model_list.json — confirmed mistral-large-latest (line 11064), mistral-medium-3.5 (line 11091), mistral-small-latest (line 11299) all lack input_cache_read_cost_per_mil_tokens
• packages/proxy/schema/model_list.json — confirmed zai-glm-5-2 (added via [BOT ISSUE] Mistral: add missing zai-glm-5-2 (Z.ai GLM 5.2) hosted model #1097) has input_cache_read_cost_per_mil_tokens: 0.14, showing the field is expected

{
  "kind": "stale_metadata",
  "provider": "mistral",
  "models": ["mistral-large-latest", "mistral-medium-3.5", "mistral-small-latest"],
  "model_specs": {
    "mistral-large-latest": {
      "format": "openai",
      "flavor": "chat",
      "multimodal": true,
      "input_cost_per_mil_tokens": 0.5,
      "output_cost_per_mil_tokens": 1.5,
      "input_cache_read_cost_per_mil_tokens": 0.05,
      "displayName": "Mistral Large",
      "max_input_tokens": 262144,
      "max_output_tokens": 262144,
      "available_providers": ["mistral"]
    },
    "mistral-medium-3.5": {
      "format": "openai",
      "flavor": "chat",
      "multimodal": true,
      "input_cost_per_mil_tokens": 1.5,
      "output_cost_per_mil_tokens": 7.5,
      "input_cache_read_cost_per_mil_tokens": 0.15,
      "displayName": "Mistral Medium 3.5",
      "reasoning": true,
      "reasoning_budget": true,
      "max_input_tokens": 256000,
      "available_providers": ["mistral"]
    },
    "mistral-small-latest": {
      "format": "openai",
      "flavor": "chat",
      "multimodal": true,
      "input_cost_per_mil_tokens": 0.15,
      "output_cost_per_mil_tokens": 0.6,
      "input_cache_read_cost_per_mil_tokens": 0.015,
      "displayName": "Mistral Small",
      "reasoning": true,
      "reasoning_budget": true,
      "max_input_tokens": 262144,
      "max_output_tokens": 131072,
      "available_providers": ["mistral"]
    }
  },
  "source_urls": [
    "https://docs.mistral.ai/inference/pricing/",
    "https://docs.mistral.ai/getting-started/models/models_overview/"
  ]
}

Summary: Add input_cache_read_cost_per_mil_tokens to all current Mistral flagship model entries. The cached input pricing follows a consistent 10× discount (10% of standard input cost) and is explicitly published on Mistral's pricing page.

[github-actions]

Prepared catalog updates for mistral-large-latest, mistral-medium-3.5, mistral-small-latest. Updated models: mistral-large-latest, mistral-medium-3.5, mistral-small-latest.

Verification sources: https://docs.mistral.ai/inference/pricing/, https://docs.mistral.ai/getting-started/models/models_overview/
Verified metadata applied:

• mistral-large-latest: format=openai, flavor=chat, displayName=Mistral Large, max_input_tokens=262144, max_output_tokens=262144, pricing(input/output per 1M)=0.5/1.5, input_cache_read_cost_per_mil_tokens=0.05, providers=mistral, multimodal=true
• mistral-medium-3.5: format=openai, flavor=chat, displayName=Mistral Medium 3.5, max_input_tokens=256000, max_output_tokens=not provided, pricing(input/output per 1M)=1.5/7.5, input_cache_read_cost_per_mil_tokens=0.15, providers=mistral, reasoning=true, reasoning_budget=true, multimodal=true
• mistral-small-latest: format=openai, flavor=chat, displayName=Mistral Small, max_input_tokens=262144, max_output_tokens=131072, pricing(input/output per 1M)=0.15/0.6, input_cache_read_cost_per_mil_tokens=0.015, providers=mistral, reasoning=true, reasoning_budget=true, multimodal=true

Included in the daily batch PR: #1173