22,153+ models across all 12 AI modalities. Two API surfaces — raw inference and intelligent inference — unlocked by one key. PAYG credits for every model (from £2e-07/M)d-source and non-chat models. Below is every feature of the openrout.ing platform, nothing omitted.
Direct OpenAI-compatible model access with zero orchestration. Every request goes straight to the model and back — no containers, no enrichment, no overhead. Maximum speed, minimum latency, every model in the catalog.
The same catalog with platform-level enrichment: context injection, auto-summarization, model aliases, coalition fallback, structured output. All platform features included free with every plan.
One API, every model: 10,210 open-weight LLMs + 25,691 PAYG models. The largest single-API catalog on Earth.
Open-weight flagships (GLM-5.3, Kimi-K3, DeepSeek-V4-Pro, Qwen3-Coder-480B) plus frontier closed-source (Claude Sonnet/Opus/Fable 5.x, GPT-6.1 Sol/Luna/Astra, Grok 4.7, Gemini 3.8 Flash).
Text-to-image, image-to-image, editing, layer decomposition. Top: Seedream 5.0 Flash £0.4651, FLUX Redux Pro, Nano Banana Pro, GPT-Image 2.5, Qwen Image 2.0 Pro Edit.
Text-to-video, image-to-video up to 4K. Top: Seedance 2.5 Intl 4K £10.20, Wan 2.2, Kling 2.5, Hailuo 02.
Text/image-to-3D mesh generation with texturing. Top: Meshy V7 £1.81, Tripo 3.0, Stable Fast 3D.
Full track generation, voice synthesis, sound design. Top: MiniMax Speech 2.6 £0.684.
Expressive text-to-speech, 30+ languages. Top: OpenAudio S1 £6.82/M, Gemini 2.5 Pro TTS £5.46/M.
Transcription, diarization, subtitles. Whisper-family, Canary, Parakeet.
Talking-head avatars, presenter videos. Multi-vendor PAYG.
Audio-driven lip synchronization for video and avatars.
Face replacement for video and image generation.
Qwen3-Embedding-0.6B (1024 dims), BGE, GTE families — for RAG and semantic search.
Upscaling, restoration, and cross-media processing utilities.
GET /v1/models with per-model context length, capabilities, concurrency cost, plan availability. Categories, featured, search.
Full drop-in replacement for the OpenAI SDK — change one base URL, keep all your code.
RPM and concurrency queue indefinitely via Redis. You only see 429 when a token/spend budget is actually exhausted — never from capacity.
41 model mappings with automatic fallback to similar-spec models when upstream is busy. 3 retries per model before failover. X-Coalition header tells you what happened.
Server-sent events streaming for chat completions, plus full non-streaming mode.
Tier-based slots from 4 (starter) to unlimited (enterprise). Bigger models consume more slots — fair multi-tenancy.
Automatic per-client, per-day, per-model accounting in Redis — visible in the dashboard, exportable via API.
POST /v1/embeddings — Qwen3-Embedding-0.6B default, 1024 dimensions.
Aliases: general → GLM-5.1, chat → Qwen3-8B, fast → DeepSeek-V4-Flash, big → Qwen2.5-72B. Or any model by full ID.
Fire parallel requests to multiple models and take the first good response — latency optimization built into the router.
Resume dropped SSE streams from the last token — mobile and flaky-network friendly.
Enable with "enrich": true — relevant context is injected into prompts before inference. Observe via X-Enrichment-Ms and X-Enriched response headers.
Long conversations are summarized automatically — context stays cheap, threads run indefinitely without token bloat.
Core (always loaded), recall (session), archival (long-term with temporal search). Automatic consolidation across tiers — zero infrastructure for you.
Upload documents and text — indexed for retrieval and structured queries via dedicated endpoints.
Configurable failover chains: if a model errors or times out, the next in your chain takes over automatically.
Per-client routing rules — route by model class, cost ceiling, or latency budget across the whole catalog.
Force schema-valid JSON responses. Plus seed reproducibility and logprobs for full determinism when you need it.
Safe retries: identical requests with the same idempotency key return the cached response instead of double-billing.
Asynchronous batch submission for non-urgent workloads at half price. DLQ + automatic retry included.
Cached embeddings billed at 20% on hit — repeated semantic queries cost a fifth.
Prioritize time-to-first-byte at 1.5× rate — for voice and interactive workloads where first token latency is everything.
Event-driven callbacks for batch completion, workflow steps, billing events — no polling required.
Six tiers from pure PAYG per-model rates (from £2e-07/M). PAYG across all 25,691 PAYG modelsen-weight LLMs. Monthly token allowance per tier.
Top-up credits for 4,988 PAYG models — frontier closed-source chat, image, video, 3D, music, TTS. Per-model per-unit pricing. Credits never expire.
Cached input tokens on PAYG chat models billed at 35% of input price — verified across all 2,425 priced chat models.
Hard daily/monthly spend ceilings per client. Alerts before you hit limits, enforcement when you do.
Card payments via Square checkout. Crypto payments with invoices, status polling, and webhooks.
Per-client, per-model cost breakdowns. See exactly where every credit goes.
15% recurring commission on all sales you refer.
JWT sessions, scoped API keys with tier limits, RPM/TPD, concurrency, model whitelists, IP whitelists, custom prefixes, expiry. One key works on both surfaces.
Team access control — grant scoped keys to teammates with per-key budgets and model access.
Interactive bot interface for account management and inference directly from chat.
Exportable audit logs of every request. Circuit breakers trip on failing upstreams and reroute automatically.
Metrics, errors, usage, snapshots, replayable event log — full visibility into platform behavior.
Per-client feature toggles with bulk and reset operations — progressively roll out platform changes to your team.
Prompts and completions are never stored, logged, or transmitted to third parties. Ephemeral by design. GDPR compliant by default.
Per-model failover chains across upstream regions and providers — resilience built into the router, not your code.
Programmatic rate-limit configuration: tiers, check, increment, cleanup — manage quotas via API.
Full OpenAPI spec at /openapi.json, downloadable Postman collection, interactive Swagger docs at api.openrout.ing/docs.
Categories, featured, search, per-model capability metadata — the catalog is fully queryable via API.
Full API reference: /docs · Pricing: /pricing · Models: /models