fix: add Databricks models databricks-gemini-3-flash +2 more

[github-actions]

fix: add Databricks models databricks-gemini-3-flash +2 more

Closes #643

Source issue: #643

Summary

Field	Value
Provider	databricks
Primary model	databricks-gemini-3-flash
Changed models	databricks-gemini-3-flash databricks-gemini-3-1-flash-lite databricks-gpt-oss-20b
Added models	databricks-gemini-3-flash databricks-gemini-3-1-flash-lite databricks-gpt-oss-20b
Updated models	None
Verification sources	1 2 3 4

Verified metadata

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
databricks-gemini-3-flash	Gemini 3 Flash		databricks	openai	chat	input=1048576, output=65536	n/a	multimodal=true
databricks-gemini-3-1-flash-lite	Gemini 3.1 Flash-Lite		databricks	openai	chat	input=1048576, output=65536	n/a	multimodal=true
databricks-gpt-oss-20b	GPT-OSS 20B		databricks	openai	chat	input=131072, output=not provided	n/a	reasoning=true

Verification notes

Verification

Official sources consulted

1. Databricks Supported models page (https://docs.databricks.com/aws/en/machine-learning/foundation-model-apis/supported-models) — Verified model existence, model IDs, supported inputs, context window for GPT-OSS 20B (128K), and reasoning capability for GPT-OSS 20B. Gemini token limits not published here.
2. Databricks Foundation model overview (https://docs.databricks.com/aws/en/machine-learning/model-serving/foundation-model-overview) — Confirmed regional availability of all three models.
3. Google AI Gemini 3 docs (https://ai.google.dev/gemini-api/docs/gemini-3) — Token limits for Gemini 3 family: gemini-3-flash-preview has 1M input / 64K output; gemini-3.1-flash-lite has 1M input / 64K output. These are the underlying models Databricks hosts.
4. HuggingFace gpt-oss-20b config.json (https://huggingface.co/openai/gpt-oss-20b) — Confirmed max_position_embeddings: 131072; architecture is MoE with 21B params / 3.6B active; supports reasoning with adjustable effort levels.

sync_models (LiteLLM) cross-check

None of the three Databricks models (databricks-gemini-3-flash, databricks-gemini-3-1-flash-lite, databricks-gpt-oss-20b) exist in the LiteLLM model_prices_and_context_window_backup.json catalog. No Databricks-prefixed models are present in sync_models at all. Therefore, no numeric field comparison is possible.

Fields not published or not applicable

• Pricing (all three models): Databricks does not publish stable per-model token pricing. Omitted.
• max_output_tokens (databricks-gpt-oss-20b): Not specified in Databricks docs. Other providers vary (131072 on Fireworks, 32768 on Groq/Together). Omitted rather than guess.
• reasoning (Gemini models): Underlying Google models support reasoning, but Databricks docs do not confirm reasoning parameter support for their Gemini endpoints. Omitted.
• reasoning_budget (GPT-OSS 20B): HuggingFace confirms adjustable reasoning effort, but Databricks docs do not confirm the reasoning_budget API parameter. Omitted.
• parent: None of these models are dated snapshots or location-scoped variants. No parent relationship applies.
• deprecation: None of these models are marked as deprecated or retiring on the Databricks docs.
• supported_regions: Not applicable — these are Databricks models, not Vertex models.

Local files inspected

• packages/proxy/schema/model_list.json — grep for databricks-gemini-3-flash, databricks-gemini-3-1-flash-lite, and databricks-gpt-oss-20b returns no matches

Source URLs

• https://docs.databricks.com/aws/en/machine-learning/model-serving/foundation-model-overview
• https://docs.databricks.com/aws/en/machine-learning/foundation-model-apis/supported-models
• https://ai.google.dev/gemini-api/docs/gemini-3
• https://huggingface.co/openai/gpt-oss-20b

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-gemini-3-flash	catalog entry	present	missing	None
databricks-gemini-3-1-flash-lite	catalog entry	present	missing	None
databricks-gpt-oss-20b	max_output_tokens	n/a	131072	databricks/databricks-gpt-oss-20b
databricks-gpt-oss-20b	input_cost_per_mil_tokens	n/a	0.07	databricks/databricks-gpt-oss-20b
databricks-gpt-oss-20b	output_cost_per_mil_tokens	n/a	0.30002	databricks/databricks-gpt-oss-20b

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
ai-proxy	Ready	Preview, Comment	May 23, 2026 6:33pm