fix: add Bedrock models minimax.minimax-m2.5

[github-actions]

fix: add Bedrock models minimax.minimax-m2.5

Closes #628

Source issue: #628

Summary

Field	Value
Provider	bedrock
Primary model	minimax.minimax-m2.5
Changed models	minimax.minimax-m2.5
Added models	minimax.minimax-m2.5
Updated models	None
Verification sources	1 2 3 4 5

Verified metadata

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
minimax.minimax-m2.5	MiniMax M2.5		bedrock	converse	chat	input=196000, output=8192	n/a	active

Verification notes

Verification checklist

• Model ID confirmed in >=2 independent official sources (model card + models-at-a-glance)
• Model ID format matches Bedrock convention (provider.model)
• Not already present in model_list.json (confirmed via grep for minimax.)
• Standard/on-demand tier confirmed
• Context window and max output tokens confirmed from AWS model card
• Bedrock-specific pricing not verified (see Verification section below)

Verification

Fields verified from official sources

Field	Value	Source
format	converse	AWS model card — Converse API supported
flavor	chat	AWS model card — Chat Completions API supported, text-in/text-out, messages-based API; matches all other Bedrock converse-format models in model_list.json
max_input_tokens	196000	AWS model card — "Context window: 196K tokens"
max_output_tokens	8192	AWS model card — "Max output tokens: 8K"
multimodal	not set (text-only)	AWS model card — input: text only, output: text only; no image/audio/video
displayName	MiniMax M2.5	AWS model card and MiniMax official site
available_providers	["bedrock"]	AWS model card — model available via bedrock-runtime and bedrock-mantle endpoints

Fields NOT verified / omitted

• Pricing (input_cost_per_mil_tokens, output_cost_per_mil_tokens): The AWS Bedrock pricing page (https://aws.amazon.com/bedrock/pricing/) lists a MiniMax AI section but the content is too long to extract programmatically. The AWS model card directs users to the pricing page but does not embed pricing. MiniMax's own API pricing ($0.30/M input, $1.20/M output from https://platform.minimax.io/docs/guides/pricing-paygo) differs from Bedrock pricing. Omitted rather than guessing.
• parent: No parent model relationship — this is the base model, not a dated snapshot or location-scoped variant.
• supported_regions: Not applicable — supported_regions is only used for Vertex models in the catalog.
• reasoning / reasoning_budget: The MiniMax marketing describes M2.5 as having reasoning capabilities, but the AWS model card does not expose reasoning/thinking token controls via the Bedrock API. Omitted.

Corrections from original issue body

• flavor changed from completion to chat: The model supports the Chat Completions API and Converse API with a messages-based interface. All existing Bedrock converse-format models in the catalog use flavor: "chat". The completion flavor is for legacy text-completion models.

sync_models (LiteLLM) cross-check

The LiteLLM model_prices_and_context_window_backup.json catalog contains no entries for any MiniMax model (searched for "minimax" in all keys). Therefore there are no sync_models values to compare against. All proposed values are sourced from the official AWS Bedrock model card.

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
minimax.minimax-m2.5	max_input_tokens	196000	1000000	minimax.minimax-m2.5
minimax.minimax-m2.5	input_cost_per_mil_tokens	n/a	0.3	minimax.minimax-m2.5
minimax.minimax-m2.5	output_cost_per_mil_tokens	n/a	1.2	minimax.minimax-m2.5

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
ai-proxy	Ready	Preview, Comment	May 22, 2026 8:43pm