API compatibility - 8 documented limits

Together AI API compatibility

Together AI documents an OpenAI-compatible API at https://api.together.ai/v1 (chat/completions, completions, embeddings, images/generations, audio/speech, audio/transcriptions, audio/translations and models). Its docs do not mention an Anthropic-compatible endpoint. The docs list 8 limitations: 8 for the OpenAI-compatible layer.

Verified October 18, 2026 against Together AI's own documentation. Not documented means the docs read that day do not say; it never means No.

Compare every provider in the API compatibility matrix.

OpenAI-compatible API

Supported

Base URL

https://api.together.ai/v1

Endpoints

chat/completions
completions
embeddings
images/generations
audio/speech
audio/transcriptions
audio/translations
models

Auth

Authorization: Bearer $TOGETHER_API_KEY

Notes

Model IDs are namespaced <provider>/<model_name>; OpenAI model strings return 404. Not supported via OpenAI shape: Assistants/Threads/Runs, fine_tuning.jobs, batches, moderations; files is partial (Together-native Files API). Docs use the api.together.ai host.

From the docs

Together's API is compatible with the OpenAI REST API and SDKs across chat, completions, vision, image generation, text-to-speech, and embeddings. ... Set api_key to your Together API key ... and base_url to https://api.together.ai/v1

docs.together.ai/docs/inference/openai-compatibility

Anthropic-compatible API

Not documented

Notes: The Together OpenAI compatibility page read today does not mention an Anthropic-compatible endpoint.

Checked: docs.together.ai/docs/inference/openai-compatibility

Documented limitations

Parameters, content types and features Together AI's docs say its compatibility layers ignore, reject or change, each with the line it comes from. A layer with nothing listed may still differ in ways the docs do not spell out.

OpenAI-compatible layer8

Assistants, Threads, Runs not implemented
Assistants, Threads, and Runs are not implemented.
docs.together.ai/docs/inference/openai-compatibility
OpenAI-shaped Batch and Files APIs not exposed
The OpenAI-shaped Batch API and Files API are not exposed through /v1.
docs.together.ai/docs/inference/openai-compatibility
moderations not implemented
moderations.create is not implemented.
docs.together.ai/docs/inference/openai-compatibility
logit_bias not supported on most models
logit_bias is not supported on most models.
docs.together.ai/docs/inference/openai-compatibility
service_tier, store, metadata, prediction ignored
service_tier, store, metadata, and prediction are accepted but ignored.
docs.together.ai/docs/inference/openai-compatibility
image_url.detail ignored
The detail field is accepted but ignored.
docs.together.ai/docs/inference/openai-compatibility
n not supported on every model
n (multiple completions per request) is supported on most chat models but not on every model.
docs.together.ai/docs/inference/openai-compatibility
logprobs returns Together's own shape
logprobs returns Together's own shape, which is richer than OpenAI's.
docs.together.ai/docs/inference/openai-compatibility

Example from Together AI's docs

Copied verbatim from the provider's documentation, placeholders included.

Provider example

import os
import openai

client = openai.OpenAI(
    api_key=os.environ.get("TOGETHER_API_KEY"),
    base_url="https://api.together.ai/v1",
)

response = client.chat.completions.create(
    model="MiniMaxAI/MiniMax-M3",
    messages=[{"role": "user", "content": "Hello!"}],
)

Newsletter

One email per week. New models, new endpoints and the compatibility changes that break a base URL, and what they mean for your stack.