fix: add Bedrock models moonshotai.kimi-k2.5

[github-actions]

fix: add Bedrock models moonshotai.kimi-k2.5

Closes #627

Source issue: #627

Summary

Field	Value
Provider	bedrock
Primary model	moonshotai.kimi-k2.5
Changed models	moonshotai.kimi-k2.5
Added models	moonshotai.kimi-k2.5
Updated models	None
Verification sources	1 2 3

Verified metadata

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
moonshotai.kimi-k2.5	Kimi K2.5		bedrock	converse	chat	input=256000, output=16384	in/out=0.6/3 per 1M	multimodal=true

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 moonshotai.kimi-k2.5 — no Bedrock entry)
• Standard/on-demand tier confirmed (model card shows Standard, Priority, Flex)
• Pricing verified: $0.60/1M input, $3.00/1M output (matches existing Azure/Vertex entry for same model)

Verification

Official sources consulted

• Bedrock model card (https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-5.html): Verified model ID moonshotai.kimi-k2.5, context window 256K, max output 16K, multimodal (text+image input, text output), Converse API + Chat Completions API support, regional availability (11 regions), lifecycle Active, launch date Jan 27 2026
• Bedrock pricing page (https://aws.amazon.com/bedrock/pricing/): Moonshot AI tab lists Kimi K2.5 Standard on-demand pricing at $0.60/1M input, $3.00/1M output
• Existing catalog entry publishers/moonshotai/models/kimi-k2.5 in model_list.json: Azure/Vertex entry has input_cost_per_mil_tokens: 0.6, output_cost_per_mil_tokens: 3, max_input_tokens: 256000, confirming consistency

Corrections from original issue

• Flavor: Changed from completion to chat. The Bedrock model card shows support for Chat Completions API and Converse API with role-based message format. All other Bedrock chat models in model_list.json use "flavor": "chat".
• Max output tokens: Specified as 16,384 (16K from model card). The original issue said 16,384 which is correct.
• Pricing: Added — not present in original issue. Confirmed at $0.60/1M input, $3.00/1M output from Bedrock pricing page.

sync_models (LiteLLM) cross-check

• moonshotai.kimi-k2.5 does not exist in the LiteLLM model_prices_and_context_window_backup.json catalog for the Bedrock provider. No bedrock-prefixed kimi entries were found at all.
• Since there is no sync_models entry, there are no numeric field deviations to document. All values come from the official Bedrock model card and pricing page.

Fields not published or not applicable

• input_cache_read_cost_per_mil_tokens: Not published by Bedrock for this model — omitted
• input_cache_write_cost_per_mil_tokens: Not published by Bedrock for this model — omitted
• reasoning: Not applicable — Kimi K2.5 is not documented as a reasoning model
• parent: Not applicable — this is the base model, not a snapshot or location-scoped variant
• deprecation_date: Not applicable — model is Active with no EOL date
• supported_regions: Not applicable — supported_regions is used for Vertex provider only, not Bedrock

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
moonshotai.kimi-k2.5	max_input_tokens	256000	262144	moonshotai.kimi-k2.5
moonshotai.kimi-k2.5	max_output_tokens	16384	262144	moonshotai.kimi-k2.5

[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 6:45pm