fix: update Mistral model metadata for mistral-small-latest

[github-actions]

fix: update Mistral model metadata for mistral-small-latest

Closes #681

Source issue: #681

Summary

Field	Value
Provider	mistral
Primary model	mistral-small-latest
Changed models	mistral-small-latest
Added models	None
Updated models	mistral-small-latest
Verification sources	1 2

Verified metadata

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
mistral-small-latest	Mistral Small		mistral	openai	chat	input=262144, output=131072	in/out=0.1/0.3 per 1M	multimodal=true

Verification notes

Verification

Sources and fields verified

• https://mistral.ai/pricing/: input_cost_per_mil_tokens (0.1), output_cost_per_mil_tokens (0.3). This is the canonical pricing page and takes precedence over model card launch pricing.
• https://docs.mistral.ai/models/model-cards/mistral-small-4-0-26-03/: max_input_tokens (262144 from "256k context window"), multimodal (true — vision/OCR capabilities documented).
• https://docs.mistral.ai/getting-started/models: Confirmed mistral-small-latest alias resolves to mistral-small-2603 (Mistral Small 4, v26.03).

Fields not published or not applicable

• max_output_tokens: Mistral documentation does not publish a separate max output token limit for this model. The model card only states "256k context window" without separating input vs output. The existing mistral-small-2603 snapshot entry in the catalog also omits max_output_tokens. Omitted from the spec rather than guessed.
• input_cache_read_cost_per_mil_tokens, input_cache_write_cost_per_mil_tokens: Not published by Mistral for this model.
• reasoning, reasoning_budget: The model card describes reasoning capabilities but Mistral does not expose a thinking/reasoning token budget via the API in the same way as other providers. Omitted.

Pricing discrepancy with original issue body

The original issue body cited $0.15/$0.60 from the model card page. However, the Mistral pricing page (mistral.ai/pricing) — the canonical reference for current API pricing — lists $0.1/$0.3. The model card likely reflects launch pricing that was subsequently reduced. The pricing page value ($0.1/$0.3) is used.

Cross-check against sync_models (LiteLLM)

Field	Proposed Value	sync_models Value	sync_models Key	Justification
max_input_tokens	262144	32000	mistral/mistral-small-latest	sync_models is stale; Mistral model card confirms 256k context window (https://docs.mistral.ai/models/model-cards/mistral-small-4-0-26-03/)
input_cost_per_mil_tokens	0.1	0.14	mistral/mistral-small-latest	sync_models is stale; Mistral pricing page shows $0.1/M (https://mistral.ai/pricing/)
output_cost_per_mil_tokens	0.3	0.42	mistral/mistral-small-latest	sync_models is stale; Mistral pricing page shows $0.3/M (https://mistral.ai/pricing/)
max_output_tokens	(omitted)	8192	mistral/mistral-small-latest	sync_models value (8192) appears stale and from an older Mistral Small version. Official docs do not publish a separate max output limit for Mistral Small 4. Omitted rather than using unverified sync_models value.

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
mistral-small-latest	max_input_tokens	262144	131072	mistral/mistral-small-latest
mistral-small-latest	input_cost_per_mil_tokens	0.1	0.06	mistral/mistral-small-latest
mistral-small-latest	output_cost_per_mil_tokens	0.3	0.18	mistral/mistral-small-latest

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
ai-proxy	Ready	Preview, Comment	May 30, 2026 10:47am