Provider Compatibility Matrix

{/This file is auto-generated by scripts/provider_compat_matrix.py. Do not edit manually./}

This matrix shows which Responses API features are supported by each inference provider, based on integration test results.

Legend	Meaning
✅	Tested and passing
❌	Tested and failing
⏭️	Skipped (unsupported)
—	Not tested

Summary

Provider	Tested	Passing	Failing	Coverage
azure	114	114	0	86%
bedrock	27	27	0	20%
ollama	2	2	0	2%
openai	133	133	0	100%
vllm	3	3	0	2%
watsonx	56	56	0	42%

Provider Details

Models, endpoints, and versions used during test recordings.

Provider	Model(s)	Endpoint	Version Info
azure	gpt-4o	llama-stack-test.openai.azure.com, lls-test.openai.azure.com	openai sdk: 2.30.0
bedrock	openai.gpt-oss-20b	bedrock-mantle.us-east-2.api.aws, bedrock-mantle.us-west-2.api.aws	openai sdk: 2.5.0
ollama	gpt-oss:20b	—	openai sdk: 2.5.0, vllm server: 0.9.2rc2.dev136+g0b382b53a.d20250924
openai	gpt-4o, o4-mini, text-embedding-3-small	api.openai.com	openai sdk: 2.5.0
vllm	Qwen/Qwen3-0.6B	—	openai sdk: 2.5.0, vllm server: 0.18.1rc1.dev197+g0e9358c11
watsonx	meta-llama/llama-3-3-70b-instruct	us-south.ml.cloud.ibm.com	openai sdk: 2.5.0

Basic Responses

Feature	azure	bedrock	ollama	openai	vllm	watsonx
extra body guided choice	✅	✅	—	✅	—	✅
include logprobs non streaming	✅	✅	—	✅	—	✅
include logprobs streaming	✅	✅	—	✅	—	✅
include logprobs with function tools	✅	✅	—	✅	—	✅
include logprobs with web search	✅	✅	—	✅	—	✅
non streaming basic	✅	✅	—	✅	—	✅
non streaming image	✅	⏭️	—	✅	—	⏭️
non streaming multi turn	✅	✅	—	✅	—	✅
non streaming multi turn image	✅	⏭️	—	✅	—	⏭️
streaming basic	✅	✅	—	✅	—	✅
streaming incremental content	✅	✅	—	✅	—	✅

Compact Responses

Feature	azure	bedrock	ollama	openai	vllm	watsonx
compact basic conversation	✅	—	—	✅	—	—
compact chain through compaction	✅	—	—	✅	—	—
compact double compaction	✅	—	—	✅	—	—
compact input items hides compaction	✅	—	—	✅	—	—
compact response is retrievable	✅	—	—	✅	—	—
compact roundtrip	✅	—	—	✅	—	—
compact single message	✅	—	—	✅	—	—
compact with previous response id	✅	—	—	✅	—	—
compact with tool calls dropped	✅	—	—	✅	—	—
context management auto compacts large input	✅	—	—	✅	—	—
context management no compact below threshold	✅	—	—	✅	—	—
context management none does not compact	✅	—	—	✅	—	—

Conversation Responses

Feature	azure	bedrock	ollama	openai	vllm	watsonx
conversation backward compatibility	✅	✅	—	✅	—	✅
conversation basic workflow	✅	✅	—	✅	—	✅
conversation context loading	✅	✅	—	✅	—	✅
conversation error handling	⏭️	⏭️	—	✅	—	⏭️
conversation multi turn and streaming	✅	✅	—	✅	—	✅

File Search

Feature	azure	bedrock	ollama	openai	vllm	watsonx
file search filter by category	✅	—	—	✅	—	⏭️
file search filter by date range	✅	—	—	✅	—	⏭️
file search filter by region	✅	—	—	✅	—	⏭️
file search filter compound and	✅	—	—	✅	—	⏭️
file search filter compound or	✅	—	—	✅	—	⏭️
file search streaming events	✅	—	—	✅	—	⏭️
text format	✅	—	—	✅	—	✅

Mcp Authentication

Feature	azure	bedrock	ollama	openai	vllm	watsonx
mcp authorization backward compatibility	✅	✅	—	✅	—	—
mcp authorization bearer	✅	✅	—	✅	—	—

Openai Responses

Feature	azure	bedrock	ollama	openai	vllm	watsonx
background false is synchronous	✅	—	—	✅	—	✅
background returns queued	✅	—	—	✅	—	✅
cancel completed response fails	✅	—	—	✅	—	⏭️
incomplete details length	✅	—	—	✅	—	✅
incomplete details length streaming	✅	—	—	✅	—	✅
incomplete details max iterations exceeded	✅	—	—	✅	—	✅
incomplete details max iterations exceeded streaming	✅	—	—	✅	—	✅
incomplete details null when completed	✅	—	—	✅	—	✅
reasoning effort	⏭️	—	—	✅	—	⏭️
reasoning effort streaming	⏭️	—	—	✅	—	⏭️
streaming includes usage	✅	—	—	✅	—	✅
streaming invalid base64 image failure code is spec compliant	✅	—	—	✅	—	⏭️
with max output tokens	✅	—	—	✅	—	⏭️
with parallel tool calls and previous response	✅	—	—	✅	—	✅
with parallel tool calls disabled	✅	—	—	✅	—	⏭️
with parallel tool calls disabled streaming	✅	—	—	✅	—	⏭️
with parallel tool calls enabled	✅	—	—	✅	—	⏭️
with prompt cache key	✅	—	—	✅	—	✅
with prompt cache key and previous response	✅	—	—	✅	—	✅
with prompt cache key streaming	✅	—	—	✅	—	✅
with safety identifier	✅	—	—	✅	—	✅
with safety identifier and previous response	✅	—	—	✅	—	✅
with safety identifier streaming	✅	—	—	✅	—	✅
with service tier	⏭️	—	—	✅	—	⏭️
with service tier and previous response	⏭️	—	—	✅	—	⏭️
with service tier auto	⏭️	—	—	✅	—	⏭️
with service tier auto and previous response	⏭️	—	—	✅	—	⏭️
with service tier auto streaming	⏭️	—	—	✅	—	⏭️
with service tier flex	⏭️	—	—	✅	—	⏭️
with service tier flex streaming	⏭️	—	—	✅	—	⏭️
with service tier streaming	⏭️	—	—	✅	—	⏭️
with small max output tokens	✅	—	—	✅	—	⏭️
with stream options and previous response	✅	—	—	✅	—	✅
with stream options includes usage	✅	—	—	✅	—	✅
with stream options non streaming	✅	—	—	✅	—	✅
with top logprobs	✅	—	—	✅	—	✅
with top logprobs and previous response	✅	—	—	✅	—	✅
with top logprobs streaming	✅	—	—	✅	—	✅
with top p	✅	—	—	✅	—	✅
with top p and previous response	✅	—	—	✅	—	✅
with top p streaming	✅	—	—	✅	—	✅
with truncation and previous response	✅	—	—	✅	—	✅
with truncation disabled	✅	—	—	✅	—	✅
with truncation disabled streaming	✅	—	—	✅	—	✅

Prompt Templates

Feature	azure	bedrock	ollama	openai	vllm	watsonx
basic prompt template	✅	✅	—	✅	—	✅
multi variable prompt template	✅	✅	—	✅	—	✅
multi version prompt template	✅	✅	—	✅	—	✅
prompt template no variables	✅	✅	—	✅	—	✅
prompt template with multi turn	✅	✅	—	✅	—	✅
prompt template with streaming	✅	✅	—	✅	—	✅

Reasoning

Feature	azure	bedrock	ollama	openai	vllm	watsonx
reasoning basic streaming	✅	✅	⏭️	✅	✅	✅
reasoning multi turn passthrough	⏭️	✅	✅	✅	✅	⏭️
reasoning non streaming	⏭️	✅	✅	✅	✅	⏭️

Responses Access Control

Feature	azure	bedrock	ollama	openai	vllm	watsonx
previous response id access control	—	—	—	✅	—	—
user can access own resources after denial	—	—	—	✅	—	—
user cannot access other users response input items	—	—	—	✅	—	—
user cannot delete other users response	—	—	—	✅	—	—
user cannot retrieve other users response	—	—	—	✅	—	—
users have isolated responses	—	—	—	✅	—	—

Responses Errors

Feature	azure	bedrock	ollama	openai	vllm	watsonx
completed response has no error	✅	✅	—	✅	—	✅
invalid base64 image returns image error	✅	⏭️	—	✅	—	⏭️
invalid image url returns image error	✅	⏭️	—	✅	—	⏭️
non vision model returns error for image input	✅	✅	—	✅	—	✅
non vision model with base64 image returns server error	✅	✅	—	✅	—	✅

Structured Output

Feature	azure	bedrock	ollama	openai	vllm	watsonx
json schema array of integers	✅	—	—	✅	—	✅
json schema array of objects	✅	—	—	✅	—	✅
json schema array of strings	✅	—	—	✅	—	⏭️
json schema boolean types	✅	—	—	✅	—	⏭️
json schema float types	✅	—	—	✅	—	⏭️
json schema integer types	✅	—	—	✅	—	⏭️
json schema mixed types structures	✅	—	—	✅	—	✅
json schema nested objects	✅	—	—	✅	—	✅
json schema string types	✅	—	—	✅	—	✅

Tool Responses

Feature	azure	bedrock	ollama	openai	vllm	watsonx
connector resolution mcp tool	✅	—	—	✅	—	—
function call ordering 1	✅	—	—	✅	—	—
function call ordering 2	✅	—	—	✅	—	—
function call output list file	✅	—	—	✅	—	—
function call output list image	✅	—	—	✅	—	—
function call output list text	✅	—	—	✅	—	—
function call output list text multi block	✅	—	—	✅	—	—
max tool calls with function tools	✅	—	—	✅	—	—
max tool calls with mcp tools	✅	—	—	✅	—	—
mcp tool approval	✅	—	—	✅	—	—
multi turn streaming web search	✅	—	—	✅	—	—
non streaming custom tool	✅	—	—	✅	—	—
non streaming file search	✅	—	—	✅	—	—
non streaming file search empty vector store	✅	—	—	✅	—	—
non streaming mcp tool	✅	—	—	✅	—	—
non streaming multi turn tool execution	✅	—	—	✅	—	—
non streaming web search	✅	—	—	✅	—	—
parallel tool calls with function tools	✅	—	—	✅	—	—
parallel tool calls with mcp tools	✅	—	—	✅	—	—
sequential file search	✅	—	—	✅	—	—
sequential mcp tool	✅	—	—	✅	—	—
streaming multi turn tool execution	✅	—	—	✅	—	—
streaming web search	✅	—	—	✅	—	—

---

Generated by scripts/provider_compat_matrix.py