Summary
The codestral-latest entry in model_list.json has stale pricing and context window values. Mistral's documentation confirms codestral-latest now points to codestral-2508, but the -latest alias retains the old Codestral 2501 pricing ($1/$3) instead of the current $0.3/$0.9.
Current values in catalog
| Field |
Current Value |
Correct Value |
input_cost_per_mil_tokens |
1 |
0.3 |
output_cost_per_mil_tokens |
3 |
0.9 |
max_input_tokens |
32000 |
256000 |
max_output_tokens |
8191 |
(remove or update) |
Evidence
The catalog's own codestral-2508 snapshot (line 5004) already has the correct values ($0.3/$0.9, 256K context), and it declares codestral-latest as its parent. This means codestral-latest is the alias for codestral-2508 but carries outdated pricing and limits from the older codestral-2501 era.
Verification
- Source 1: Mistral pricing page — lists "Codestral" at $0.3/M input, $0.9/M output
- Source 2: Mistral models documentation — confirms
codestral-latest points to codestral-25-08 (i.e. codestral-2508)
- Internal consistency: The
codestral-2508 entry in the same catalog file already has input_cost_per_mil_tokens: 0.3, output_cost_per_mil_tokens: 0.9, max_input_tokens: 256000
- ID format:
codestral-latest matches existing Mistral -latest alias convention
- Recent commits: No recent commits updating
codestral-latest pricing
Suggested fix
Update codestral-latest in packages/proxy/schema/model_list.json (currently around line 4985):
"codestral-latest": {
"format": "openai",
"flavor": "chat",
"input_cost_per_mil_tokens": 0.3,
"output_cost_per_mil_tokens": 0.9,
"displayName": "Codestral",
"max_input_tokens": 256000,
"available_providers": [
"mistral"
]
}Local files inspected
packages/proxy/schema/model_list.json — codestral-latest at line 4985 ($1/$3, 32K), codestral-2508 at line 5004 ($0.3/$0.9, 256K)- No existing open issue covers this model's pricing
{
"kind": "cost_update",
"provider": "mistral",
"models": ["codestral-latest"],
"status": "active",
"fields_to_update": {
"input_cost_per_mil_tokens": {"old": 1, "new": 0.3},
"output_cost_per_mil_tokens": {"old": 3, "new": 0.9},
"max_input_tokens": {"old": 32000, "new": 256000},
"max_output_tokens": {"old": 8191, "new": null}
},
"source_urls": [
"https://mistral.ai/pricing",
"https://docs.mistral.ai/getting-started/models/"
]
}
Summary
The
codestral-latestentry inmodel_list.jsonhas stale pricing and context window values. Mistral's documentation confirmscodestral-latestnow points tocodestral-2508, but the-latestalias retains the old Codestral 2501 pricing ($1/$3) instead of the current $0.3/$0.9.Current values in catalog
input_cost_per_mil_tokensoutput_cost_per_mil_tokensmax_input_tokensmax_output_tokensEvidence
The catalog's own
codestral-2508snapshot (line 5004) already has the correct values ($0.3/$0.9, 256K context), and it declarescodestral-latestas its parent. This meanscodestral-latestis the alias forcodestral-2508but carries outdated pricing and limits from the oldercodestral-2501era.Verification
codestral-latestpoints tocodestral-25-08(i.e.codestral-2508)codestral-2508entry in the same catalog file already hasinput_cost_per_mil_tokens: 0.3,output_cost_per_mil_tokens: 0.9,max_input_tokens: 256000codestral-latestmatches existing Mistral-latestalias conventioncodestral-latestpricingSuggested fix
Update
codestral-latestinpackages/proxy/schema/model_list.json(currently around line 4985):Local files inspected
packages/proxy/schema/model_list.json—codestral-latestat line 4985 ($1/$3, 32K),codestral-2508at line 5004 ($0.3/$0.9, 256K){ "kind": "cost_update", "provider": "mistral", "models": ["codestral-latest"], "status": "active", "fields_to_update": { "input_cost_per_mil_tokens": {"old": 1, "new": 0.3}, "output_cost_per_mil_tokens": {"old": 3, "new": 0.9}, "max_input_tokens": {"old": 32000, "new": 256000}, "max_output_tokens": {"old": 8191, "new": null} }, "source_urls": [ "https://mistral.ai/pricing", "https://docs.mistral.ai/getting-started/models/" ] }