[BOT ISSUE] Databricks: add missing databricks-gpt-6-astra, databricks-glm-5-3, databricks-deepseek-v4-1-flash

[github-actions]

Summary

Three chat/completion models are listed on Databricks Foundation Model APIs' supported models page (updated Sep 15, 2026) but are absent from model_list.json as Databricks-specific entries. The underlying models already exist in the catalog under other providers.

Missing Models

1. databricks-gpt-6-astra

Field	Expected value	Source
format	"openai"	matches base gpt-6-astra (line 227)
flavor	"chat"	matches base
multimodal	true	matches base
displayName	"GPT-6 Astra"	matches base
reasoning	true	matches base
max_input_tokens	922000	matches base
max_output_tokens	128000	matches base
available_providers	["databricks"]	standard Databricks pattern

Base model: gpt-6-astra (line 227), currently available on openai, azure, openrouter.

2. databricks-glm-5-3

Field	Expected value	Source
format	"openai"	matches sibling databricks-glm-5-3-flash (line 13399)
flavor	"chat"	matches sibling
displayName	"GLM 5.3"	matches base zai-org/GLM-5.3 (line 11438)
reasoning	true	matches sibling and base
max_input_tokens	1048576	matches sibling and base
max_output_tokens	131072	matches sibling (base has 262144; verify with Databricks)
available_providers	["databricks"]	standard Databricks pattern

Sibling databricks-glm-5-3-flash (line 13399) already exists. Base model: zai-org/GLM-5.3 (line 11438), available on baseten, together. Databricks pricing page confirms GLM-5.3 at 2.143/7.143 DBU/MTok (distinct from GLM-5.3 Flash at the same rates).

3. databricks-deepseek-v4-1-flash

Field	Expected value	Source
format	"openai"	matches sibling databricks-deepseek-v4-flash-0731 (line 13432)
flavor	"chat"	matches sibling
multimodal	true	matches base deepseek-ai/DeepSeek-V4.1-Flash (line 11385)
displayName	"DeepSeek V4.1 Flash"	matches base
reasoning	true	matches sibling and base
max_input_tokens	1048576	matches sibling and base
available_providers	["databricks"]	standard Databricks pattern

Sibling databricks-deepseek-v4-flash-0731 (line 13432) already exists. Base model: deepseek-ai/DeepSeek-V4.1-Flash (line 11385), available on baseten, fireworks, together.

Verification checklist

Check	Status	Detail
Cross-source	✅	All three confirmed on Databricks supported models docs (updated Sep 15, 2026). databricks-glm-5-3 also confirmed on Databricks pricing page (2.143/7.143 DBU). Underlying models verified in catalog under other providers
Not in catalog	✅	None of the three Databricks-specific model IDs exist in model_list.json
ID format	✅	Follows existing databricks-{model} convention (e.g., databricks-glm-5-3-flash, databricks-deepseek-v4-flash-0731)
Not deprecated	✅	All three are active on the Databricks supported models page as of Sep 15, 2026
Not duplicate	✅	#1092 covers databricks-gpt-5-4-mini and databricks-gemini-3-6-flash — different models
Not embedding/CV-only	✅	All three are chat/completion models

Sources

• Databricks Foundation Model APIs supported models: https://docs.databricks.com/en/machine-learning/foundation-models/supported-models.html (updated Sep 15, 2026)
• Databricks pay-per-token pricing: https://www.databricks.com/product/pricing/foundation-model-serving (confirms GLM 5.3 availability)

[github-actions]

Prepared catalog updates for databricks-gpt-6-astra, databricks-glm-5-3, databricks-deepseek-v4-1-flash. Added models: databricks-gpt-6-astra, databricks-glm-5-3, databricks-deepseek-v4-1-flash.

Verification sources: https://docs.databricks.com/en/machine-learning/foundation-models/supported-models.html, https://www.databricks.com/product/pricing/foundation-model-serving
Verified metadata applied:

• databricks-gpt-6-astra: format=openai, flavor=chat, displayName=GPT-6 Astra, max_input_tokens=922000, max_output_tokens=128000, pricing=not provided, providers=databricks, reasoning=true, multimodal=true
• databricks-glm-5-3: format=openai, flavor=chat, displayName=GLM 5.3, max_input_tokens=1048576, max_output_tokens=65536, pricing=not provided, providers=databricks, reasoning=true
• databricks-deepseek-v4-1-flash: format=openai, flavor=chat, displayName=DeepSeek V4.1 Flash, max_input_tokens=1048576, max_output_tokens=not provided, pricing=not provided, providers=databricks, reasoning=true, multimodal=true

Included in the daily batch PR: #1277