One endpoint.
Every model.
Chat, vision, image, video, embeddings, transcription, speech and rerank behind a single OpenAI-compatible API. Standardized parameters, streaming everywhere, and the provider layer managed entirely by APIRaven.
- 749
- models in the catalog
- 9
- modalities behind one key
- OpenAI
- compatible request shape
- 1
- balance, no provider accounts
https://api.apiraven.com/v1curl https://api.apiraven.com/v1/chat/completions \
-H "Authorization: Bearer $APIRAVEN_API_KEY" \
-d '{
"model": "openai/gpt-oss-120b",
"messages": [{ "role": "user", "content": "Hello" }],
"stream": true
}'One API, five modalities
Every model speaks the same request and response shape — switch tasks by changing one string, not your integration.
Chat models
Text in, text out — streaming, tools and JSON mode.
219 models →
Vision models
Text and images in, text out.
213 models →
Image models
Text to image generation.
103 models →
Video models
Text to video generation.
57 models →
Embedding models
Text to vectors for search and RAG.
59 models →
Transcription models
Speech to text with timestamps.
28 models →
Speech models
Text to speech synthesis.
48 models →
Rerank models
Cross-encoder reranking for retrieval.
10 models →
Audio models
Real-time speech to speech and voice.
12 models →
The model catalog
Search by name, filter by modality and capability, then open a model for parameters, pricing and copy-paste code.
749 models
One key · OpenAI-compatible · provider managed
Nova 2 Lite
amazon/nova-2-lite
- Context
- 1M
- Max output
- 1M
- Speed
- 0 tok/s
Nova Lite
amazon/nova-lite
- Context
- 300K
- Max output
- 8K
- Speed
- 0 tok/s
Nova Micro
amazon/nova-micro
- Context
- 128K
- Max output
- 8K
- Speed
- 0 tok/s
Nova Premier
amazon/nova-premier
- Context
- 1M
- Max output
- 0
- Speed
- 0 tok/s
Nova Pro
amazon/nova-pro
- Context
- 300K
- Max output
- 8K
- Speed
- 0 tok/s
Titan Text Embeddings V2
amazon/titan-embed-text-2
- Context
- 0
- Dimensions
- 0
- Batch
- 2,048 texts
Claude 3 Haiku
anthropic/claude-3-haiku
- Context
- 200K
- Max output
- 4K
- Speed
- 0 tok/s
Claude 4.5 Sonnet
anthropic/claude-4.5-sonnet
- Context
- 200K
- Max output
- 0
- Speed
- 0 tok/s
Claude 4.6 Sonnet
anthropic/claude-4.6-sonnet
- Context
- 200K
- Max output
- 8K
- Speed
- 0 tok/s
Claude 5 Sonnet
anthropic/claude-5-sonnet
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude 5.5 Haiku
anthropic/claude-5.5-haiku
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Fable 5
anthropic/claude-fable-5
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Fable 5.1
anthropic/claude-fable-5.1
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Fable Latest
anthropic/claude-fable-latest
- Context
- 1M
- Max output
- 0
- Speed
- 0 tok/s
Claude Haiku 4.5
anthropic/claude-haiku-4.5
- Context
- 200K
- Max output
- 64K
- Speed
- 0 tok/s
Claude Haiku 5.5
anthropic/claude-haiku-5.5
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Haiku Latest
anthropic/claude-haiku-latest
- Context
- 1M
- Max output
- 0
- Speed
- 0 tok/s
Claude Opus 4
anthropic/claude-opus-4
- Context
- 200K
- Max output
- 32K
- Speed
- 0 tok/s
Claude Opus 4.1
anthropic/claude-opus-4.1
- Context
- 200K
- Max output
- 0
- Speed
- 0 tok/s
Claude Opus 4.5
anthropic/claude-opus-4.5
- Context
- 200K
- Max output
- 64K
- Speed
- 0 tok/s
Claude Opus 4.6
anthropic/claude-opus-4.6
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Opus 4.7
anthropic/claude-opus-4.7
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Opus 4.8
anthropic/claude-opus-4.8
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Opus 5
anthropic/claude-opus-5
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Opus 5.5
anthropic/claude-opus-5.5
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Opus Latest
anthropic/claude-opus-latest
- Context
- 1M
- Max output
- 0
- Speed
- 0 tok/s
Claude Sonnet 4
anthropic/claude-sonnet-4
- Context
- 1M
- Max output
- 64K
- Speed
- 0 tok/s
Claude Sonnet 4.5
anthropic/claude-sonnet-4.5
- Context
- 1M
- Max output
- 64K
- Speed
- 0 tok/s
Claude Sonnet 4.6
anthropic/claude-sonnet-4.6
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Sonnet 5
anthropic/claude-sonnet-5
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Sonnet 5.5
anthropic/claude-sonnet-5.5
- Context
- 1M
- Max output
- 128K
- Speed
- 0 tok/s
Claude Sonnet Latest
anthropic/claude-sonnet-latest
- Context
- 1M
- Max output
- 0
- Speed
- 0 tok/s
Arcee Trinity Large Thinking
arcee-ai/arcee-trinity-large-thinking
- Context
- 128K
- Max output
- 32K
- Speed
- 0 tok/s
Trinity Large Thinking
arcee-ai/trinity-large-thinking
- Context
- 262K
- Max output
- 80K
- Speed
- 0 tok/s
bge-base-en v1.5
baai/bge-base-en-v1.5
- Context
- 512
- Dimensions
- 0
- Batch
- 2,048 texts
bge-en ICL
baai/bge-en-icl
- Context
- 0
- Dimensions
- 0
- Batch
- 2,048 texts
bge-large-en v1.5
baai/bge-large-en-v1.5
- Context
- 512
- Dimensions
- 0
- Batch
- 2,048 texts
bge-m3
baai/bge-m3
- Context
- 8K
- Dimensions
- 0
- Batch
- 2,048 texts
bge-m3-multi
baai/bge-m3-multi
- Context
- 0
- Dimensions
- 0
- Batch
- 2,048 texts
bge-m3-multi-8k
baai/bge-m3-multi-8k
- Context
- 0
- Dimensions
- 0
- Batch
- 2,048 texts
bge-reranker v2 m3
baai/bge-reranker-v2-m3
- Documents
- 1,000
- Top N
- configurable
- Output
- scores
FLUX.1 Dev
black-forest-labs/flux-1-dev
- Speed
- 4–8 s
- Sizes
- 0 options
- Batch
- up to 4
FLUX.1 Redux Dev
black-forest-labs/flux-1-redux-dev
- Speed
- 4–8 s
- Sizes
- 0 options
- Batch
- up to 4
FLUX.1 Schnell
black-forest-labs/flux-1-schnell
- Speed
- 4–8 s
- Sizes
- 0 options
- Batch
- up to 4
FLUX.1.1 Pro
black-forest-labs/flux-1.1-pro
- Speed
- 4–8 s
- Sizes
- 0 options
- Batch
- up to 4
FLUX.2 Dev
black-forest-labs/flux-2-dev
- Speed
- 4–8 s
- Sizes
- 0 options
- Batch
- up to 4
FLUX.2 Flex
black-forest-labs/flux-2-flex
- Speed
- 4–8 s
- Sizes
- 0 options
- Batch
- up to 4
FLUX.2 Klein 4B
black-forest-labs/flux-2-klein-4b
- Speed
- 4–8 s
- Sizes
- 0 options
- Batch
- up to 4
One request shape, every backend
The catalog hides a fleet of inference backends. Requests and responses are normalized, so your code never branches on the model behind the model.
Automatic failover
Requests retry and reroute before the first token lands. One stable error envelope when a model is degraded.
Provider-free contract
Model ids are APIRaven-owned. No provider accounts, region pickers or keys to juggle.
One balance, one meter
Tokens, images and audio minutes all bill from the same APIRaven balance with normalized usage in every response.
Same SDK you already use
Point the official OpenAI SDK at our base URL and call any model in the catalog.
Upstream quirks → unified parameters
max_tokensmax_tokens · max_completion_tokens · max_new_tokens
One field. Aliases collapse to max_tokens.
stopstop · stop_sequences · stop_token_ids
Strings or arrays, one shape.
tool callsThree incompatible tool-call formats
OpenAI tool_calls in every response.
streamingEach provider ships its own SSE envelope
chat.completion.chunk, byte-for-byte consistent.
image inputBase64, URLs and provider file ids
image_url with detail: low | high | auto.
usageToken counters named differently everywhere
prompt / completion / total, always present.
errorsProvider-specific error bodies
One envelope with a stable error code.
samplingtop_k · min_p · repetition_penalty on some models
Accepted globally, mapped or dropped safely.
Open-model pricing, without the provider paperwork
Model rates are listed per 1M tokens, per image or per audio minute. Tokens are metered from the same balance you already use across APIRaven — no monthly platform fee and no per-provider minimums.
See platform pricing- $0 in·$0 out / 1MClaude 4.5 Sonnetanthropic/claude-4.5-sonnet
- $0 in·$0 out / 1MClaude 4.6 Sonnetanthropic/claude-4.6-sonnet
- $0 in·$0 out / 1MClaude 5 Sonnetanthropic/claude-5-sonnet
Unified inference, answered
Which provider actually runs my request?+
APIRaven routes each request across a pool of managed inference backends. The model id, request shape and response are identical no matter which backend serves it — you never manage provider accounts, keys or regions.
Is the API OpenAI-compatible?+
Yes. Point any OpenAI SDK at https://api.apiraven.com/v1 with your APIRaven key. Chat, vision, images, embeddings and audio all follow the OpenAI request and response shapes.
What happens when a backend degrades?+
Requests are retried and failed over automatically before the first token is returned. If a model is temporarily unavailable, the API returns one stable error envelope instead of a provider-specific failure.
How is usage billed?+
Per token for text, per image for generation and per minute for transcription — deducted from your APIRaven balance. Every response includes normalized usage so you can meter cost per request.
Can I use the same key for every model?+
Yes. One key works across the whole catalog; scopes, spend caps and model allowlists are managed per key in the dashboard.
Point your SDK at one base URL
Create a key, call any model in the catalog, and let APIRaven deal with the inference fleet behind it.