docs(snowflake): add REST API pricing for 18 Snowflake Cortex models#25707
docs(snowflake): add REST API pricing for 18 Snowflake Cortex models#25707sfc-gh-ridalemkaalel wants to merge 1 commit intoBerriAI:mainfrom
Conversation
models
Adds pricing entries for 18 Snowflake models from Table 6(b) of the
Snowflake Service Consumption Table (REST API with Prompt Caching),
using Regional pricing throughout.
- 9 Claude models (AWS Regional): claude-3-7-sonnet, claude-4-opus,
claude-4-sonnet, claude-haiku-4-5, claude-opus-4-5, claude-opus-4-6,
claude-sonnet-4-5, claude-sonnet-4-5-long-context, claude-sonnet-4-6
- 9 OpenAI models (Azure Regional): openai-gpt-4.1, openai-gpt-5,
openai-gpt-5-mini, openai-gpt-5-nano, openai-gpt-5.1, openai-gpt-5.2,
openai-gpt-5.4, openai-gpt-5.4-long-context, openai-o4-mini
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
Greptile SummaryThis PR adds pricing entries for 18 Snowflake Cortex REST API models (9 Claude via AWS Regional, 9 OpenAI via Azure Regional) to both Confidence Score: 4/5Safe to merge after addressing the missing supports_prompt_caching flag on OpenAI models One P2 data inconsistency: the 9 OpenAI Snowflake models have cache_read_input_token_cost but omit supports_prompt_caching: true, causing litellm.supports_prompt_caching() to return False for them. This is a real API-correctness gap. A second P2 asks for confirmation that gpt-5 and gpt-5.1 identical pricing is intentional. Both JSON files are in sync and the test file is well-structured. model_prices_and_context_window.json — the 9 OpenAI Snowflake model entries need supports_prompt_caching: true added
|
| Filename | Overview |
|---|---|
| model_prices_and_context_window.json | Adds 18 new snowflake model entries; OpenAI models include cache_read_input_token_cost but omit supports_prompt_caching: true, inconsistent with the existing pattern for OpenAI models |
| litellm/model_prices_and_context_window_backup.json | Kept in sync with root JSON; same 18 entries added identically |
| tests/test_litellm/llms/snowflake/test_snowflake_pricing.py | New test file; reads local JSON (no network calls), covers existence, provider, pricing, caching fields, and context window for all 18 models |
Flowchart
%%{init: {'theme': 'neutral'}}%%
flowchart TD
A[Snowflake Cortex REST API] --> B{Model Family}
B --> C[Claude Models x9\nAWS Regional]
B --> D[OpenAI Models x9\nAzure Regional]
C --> E[input_cost_per_token\noutput_cost_per_token\ncache_creation_input_token_cost\ncache_read_input_token_cost\nsupports_prompt_caching: true]
D --> F[input_cost_per_token\noutput_cost_per_token\ncache_read_input_token_cost\n⚠️ supports_prompt_caching: missing]
E --> G[model_prices_and_context_window.json\n+ backup]
F --> G
Reviews (1): Last reviewed commit: "feat(snowflake): add REST API pricing fo..." | Re-trigger Greptile
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| "snowflake/openai-gpt-4.1": { | ||
| "litellm_provider": "snowflake", | ||
| "max_input_tokens": 1047576, | ||
| "max_output_tokens": 32768, | ||
| "max_tokens": 32768, | ||
| "mode": "chat", | ||
| "input_cost_per_token": 2.2e-06, | ||
| "output_cost_per_token": 8.8e-06, | ||
| "cache_read_input_token_cost": 5.5e-07, | ||
| "supports_function_calling": true, | ||
| "supports_vision": true | ||
| }, |
There was a problem hiding this comment.
supports_prompt_caching flag on OpenAI Snowflake models
All 9 snowflake/openai-* entries carry cache_read_input_token_cost but none set "supports_prompt_caching": true. Every direct OpenAI model with cache pricing (e.g. gpt-4o-mini) includes this flag, and litellm.supports_prompt_caching("snowflake/openai-gpt-5") will return False as a result. The same pattern applies to all 9 OpenAI Snowflake entries.
| "snowflake/openai-gpt-4.1": { | |
| "litellm_provider": "snowflake", | |
| "max_input_tokens": 1047576, | |
| "max_output_tokens": 32768, | |
| "max_tokens": 32768, | |
| "mode": "chat", | |
| "input_cost_per_token": 2.2e-06, | |
| "output_cost_per_token": 8.8e-06, | |
| "cache_read_input_token_cost": 5.5e-07, | |
| "supports_function_calling": true, | |
| "supports_vision": true | |
| }, | |
| "snowflake/openai-gpt-4.1": { | |
| "litellm_provider": "snowflake", | |
| "max_input_tokens": 1047576, | |
| "max_output_tokens": 32768, | |
| "max_tokens": 32768, | |
| "mode": "chat", | |
| "input_cost_per_token": 2.2e-06, | |
| "output_cost_per_token": 8.8e-06, | |
| "cache_read_input_token_cost": 5.5e-07, | |
| "supports_prompt_caching": true, | |
| "supports_function_calling": true, | |
| "supports_vision": true | |
| }, |
| "snowflake/openai-gpt-5": { | ||
| "litellm_provider": "snowflake", | ||
| "max_input_tokens": 272000, | ||
| "max_output_tokens": 128000, | ||
| "max_tokens": 128000, | ||
| "mode": "chat", | ||
| "input_cost_per_token": 1.38e-06, | ||
| "output_cost_per_token": 1.1e-05, | ||
| "cache_read_input_token_cost": 1.4e-07, | ||
| "supports_function_calling": true, | ||
| "supports_vision": true, | ||
| "supports_reasoning": true | ||
| }, |
There was a problem hiding this comment.
openai-gpt-5 and openai-gpt-5.1
snowflake/openai-gpt-5 and snowflake/openai-gpt-5.1 share exactly the same input_cost_per_token ($1.38e-06), output_cost_per_token ($1.1e-05), and cache_read_input_token_cost ($1.4e-07). Is this intentional per Snowflake's pricing table, or a copy-paste error for one of the two entries? Are openai-gpt-5 and openai-gpt-5.1 genuinely the same price tier in Table 6(b), or should one of the entries have different values?
sfc-gh-ridalemkaalel
changed the title
1 participant
Summary
claude-3-7-sonnet,claude-4-opus,claude-4-sonnet,claude-haiku-4-5,claude-opus-4-5,claude-opus-4-6,claude-sonnet-4-5,claude-sonnet-4-5-long-context,claude-sonnet-4-6— all includecache_creation_input_token_costandcache_read_input_token_costopenai-gpt-4.1,openai-gpt-5,openai-gpt-5-mini,openai-gpt-5-nano,openai-gpt-5.1,openai-gpt-5.2,openai-gpt-5.4,openai-gpt-5.4-long-context,openai-o4-mini— all includecache_read_input_token_cost(no cache write cost per the table)model_prices_and_context_window.jsonandlitellm/model_prices_and_context_window_backup.jsonare updated in sync (backup brought in line with root, which is whatci_cd/check_files_match.pyenforces)Test plan
tests/test_litellm/llms/snowflake/test_snowflake_pricing.pywith 6 tests:test_snowflake_models_exist— all 18 models present in the JSONtest_snowflake_models_have_correct_provider—litellm_provider == "snowflake"for alltest_snowflake_models_have_positive_pricing—input_cost_per_tokenandoutput_cost_per_tokenare positivetest_snowflake_claude_models_have_prompt_caching_fields— Claude models have both cache write and read costs, andsupports_prompt_caching: truetest_snowflake_openai_models_have_cache_read_but_no_cache_write— OpenAI models have cache read cost but no cache write costtest_snowflake_models_have_context_window— all models definemax_input_tokens,max_output_tokens,max_tokenspython3 -c "import json; json.load(open(...))"blackformatting check andrufflinting.... Generated with Cortex Code