fix: add Databricks models databricks-claude-sonnet-4-5 +2 more

[github-actions]

fix: add Databricks models databricks-claude-sonnet-4-5 +2 more

Closes #592

Source issue: #592

Summary

Field	Value
Provider	databricks
Primary model	databricks-claude-sonnet-4-5
Changed models	databricks-claude-sonnet-4-5 databricks-claude-opus-4-6 databricks-claude-sonnet-4
Added models	databricks-claude-sonnet-4-5 databricks-claude-opus-4-6 databricks-claude-sonnet-4
Updated models	None
Verification sources	1 2

Verified metadata

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
databricks-claude-sonnet-4-5	Claude 4.5 Sonnet		databricks	openai	chat	input=200000, output=64000	n/a	multimodal=true; reasoning=true
databricks-claude-opus-4-6	Claude 4.6 Opus		databricks	openai	chat	input=1000000, output=128000	n/a	multimodal=true; reasoning=true
databricks-claude-sonnet-4	Claude 4 Sonnet		databricks	openai	chat	input=1000000, output=64000	n/a	multimodal=true; reasoning=true

Verification notes

Verification

Official sources consulted

1. Azure Databricks foundation model overview (https://learn.microsoft.com/en-us/azure/databricks/machine-learning/model-serving/foundation-model-overview, updated 2026-05-20):

   • Confirmed all three model endpoint names exist: databricks-claude-sonnet-4-5, databricks-claude-opus-4-6, databricks-claude-sonnet-4
   • Confirmed regional availability across multiple Azure regions
   • No token limits or pricing published on this page

2. AWS Databricks supported models (https://docs.databricks.com/aws/en/machine-learning/foundation-model-apis/supported-models):

   • Confirmed all three endpoint names and that they are available on the public Foundation Model APIs surface (pay-per-token)
   • databricks-claude-sonnet-4-5: 200K token context window, text+image inputs, extended thinking support
   • databricks-claude-opus-4-6: 1 million token context window, text+image inputs, adaptive thinking support
   • databricks-claude-sonnet-4: text+image inputs, extended thinking support; context window not explicitly stated
   • Max output tokens and pricing not documented for any of the three models

3. Anthropic base model reference (from existing model_list.json entries):

   • claude-sonnet-4-5: max_input=200000, max_output=64000, multimodal=true, reasoning=true, reasoning_budget=true
   • claude-opus-4-6: max_input=1000000, max_output=128000, multimodal=true, reasoning=true, reasoning_budget=true
   • claude-sonnet-4-20250514 (Claude Sonnet 4): max_input=1000000, max_output=64000, multimodal=true, reasoning=true, reasoning_budget=true

Fields verified

Field	Source	Notes
format: "openai"	Existing databricks-claude-3-7-sonnet entry	Databricks Foundation Model APIs use OpenAI-compatible format
flavor: "chat"	Existing databricks-claude-3-7-sonnet entry	Chat completion endpoint
multimodal: true	AWS Databricks docs (text+image inputs)	Confirmed for all three models
reasoning: true	AWS Databricks docs (extended/adaptive thinking)	Confirmed for all three models
reasoning_budget: true	Anthropic base models	All underlying models support reasoning budgets
max_input_tokens	Databricks docs + Anthropic base models	Sonnet 4.5: 200K confirmed by Databricks; Opus 4.6: 1M confirmed by Databricks; Sonnet 4: 1M from Anthropic base (not explicitly stated in Databricks docs)
max_output_tokens	Anthropic base models	Not documented by Databricks; using Anthropic values (64K for Sonnet models, 128K for Opus)
displayName	Anthropic base model displayName convention	Matches existing pattern: "Claude 4.5 Sonnet", "Claude 4.6 Opus", "Claude 4 Sonnet"
available_providers	Databricks docs	["databricks"] — confirmed as public serverless endpoints

Fields omitted

• Pricing (input_cost_per_mil_tokens, output_cost_per_mil_tokens): Databricks does not publish per-token pricing for these Claude models in their public documentation. The existing databricks-claude-3-7-sonnet entry has pricing (2.99999/15.00002), but no equivalent pricing source was found for the new models. Omitted rather than guessed.
• Cache pricing: Not applicable to Databricks provider.
• parent: These are not dated snapshots, location-scoped variants, or versioned aliases of another model in model_list.json. They are Databricks-specific endpoint names for their serverless API. No parent relationship applies.
• supported_regions: Not applicable — available_providers does not include "vertex".
• locations: Not required — these are Databricks endpoints, not Vertex publishers/ paths.

sync_models (LiteLLM) cross-check

No entries matching databricks-claude-sonnet-4-5, databricks-claude-opus-4-6, or databricks-claude-sonnet-4 exist in the LiteLLM model_prices_and_context_window_backup.json catalog. Therefore, no numeric field comparison is possible. All proposed values are sourced from official Databricks documentation and Anthropic base model entries in model_list.json.

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
databricks-claude-sonnet-4-5	input_cost_per_mil_tokens	n/a	2.99999	databricks/databricks-claude-sonnet-4-5
databricks-claude-sonnet-4-5	output_cost_per_mil_tokens	n/a	15.00002	databricks/databricks-claude-sonnet-4-5
databricks-claude-opus-4-6	catalog entry	present	missing	None
databricks-claude-sonnet-4	max_input_tokens	1000000	200000	databricks/databricks-claude-sonnet-4
databricks-claude-sonnet-4	input_cost_per_mil_tokens	n/a	2.99999	databricks/databricks-claude-sonnet-4
databricks-claude-sonnet-4	output_cost_per_mil_tokens	n/a	15.00002	databricks/databricks-claude-sonnet-4

[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 12:55am