01
Genuine models, nothing re-wrapped
Every model is the real thing — full context window, native tools, native vision. No template proxies, no quietly downgraded variants, no truncated context behind your back.
First-hand · Traceable to AWS Bedrock / VertexFirst-hand models.
First-hand models.
Traceable to the source.
Real Claude and Gemini, sourced through tier-1 cloud channels you can trace — AWS Bedrock, Google Vertex — never a gray-market pool. One OpenAI-compatible endpoint, priced below official list. Swap one base_url.
Drop-in OpenAI SDK · Change one base URL · You’re live
Use Brievio as your model provider — an OpenAI-compatible gateway to every first-party text, image and video model. base_url: https://api.brievio.com/v1 auth: Authorization: Bearer $BRIEVIO_API_KEY To use a model, call GET /v1/models for the live catalog, then route each model by its brievio.endpoint field. Full agent reference: https://brievio.com/llms.txt
prompt_tokens
14
completion_tokens
9
cost_usd
$0.000174
$Built on first-party-grade infrastructure
- 16+Models in the catalog
- 7First-party chat models
- ~15%Under official list
- 99.95%Measured uptime
- HonestToken billing
- Drop-inOpenAI SDK
cat ./why-brievio.md
Genuine models on infrastructure you can build a business on.
Brievio is the discounted-official tier: the real Claude, Gemini and top image/video models, served on enterprise-grade backends, billed on honest token counts. Reliability and authenticity first — never a race to the bottom on price.
02
OpenAI-compatible, drop-in
Keep the OpenAI SDK you already wrote. Streaming, function calling, tool use and vision all behave exactly as upstream — point base_url at https://api.brievio.com/v1 and ship.
03
Reliability you can build on
Requests complete fast, or fail loud and fast so your retries actually work. No 90-second hangs, no silent rate-walls — automatic failover the moment a backend degrades.
04
A fair price, not a fire sale
Roughly 15% under each provider’s official list, per model — and top-up bonuses push the effective discount to about 21%. We’re deliberately not the cheapest endpoint online — the 80%-off ones resell gray-market capacity that disappears overnight.
05
Billing you can audit
True token counts straight from the model, never padded by hidden system prompts. Every request is logged with real input/output tokens and exact cost. Failed requests are never billed.
06
Monitored, fail-fast routing
Health is watched continuously. When an upstream wobbles, traffic reroutes before your users feel it — and when something does break, it fails fast instead of hanging.
07
Native streaming, real tokens
Server-sent events passed straight through. Time-to-first-token tracks the upstream provider — no buffering, no batching, no synthetic delay inserted in the middle.
08
Every call accounted for
Per-call analytics by model, key and IP, with the genuine token counts behind each charge. Export the full ledger as CSV whenever finance asks.
09
Prompt caching, honored natively
Where the provider supports it, cache_control on your system prompt is passed straight through — real cache hits, real savings, with hit rate and saved spend shown live in your dashboard.
ls ./use-cases
Browse all use casesWhat to build with Brievio.
- Customer Support
AI customer support
The fastest-ROI AI deployment in any B2C SaaS — automate ticket triage, draft 80% of responses, and escalate the rest cleanly. Production code, real cost numbers, and the compliance pitfalls that catch teams off-guard.
Explore - Knowledge Base
RAG chatbot API
Most internal knowledge bases are dead documentation — nobody finds anything. A Claude-backed RAG chatbot turns them into a real assistant that cites sources and refuses when it doesn't know. Here's the production pattern.
Explore - Trust & Safety
AI content moderation
Modern moderation isn't just regex — it's nuance: sarcasm, dog whistles, brand-context misuse, image+text combinations. LLMs do this far better than rule-based systems, at a price that scales.
Explore - Developer Tools
AI code assistant
Cursor, Aider, Cline, Continue.dev — they're all powered by the same handful of first-party LLMs. If you're building a coding tool (or a co-pilot inside your own dev product), here's the architecture and the cost reality.
Explore - Data Processing
AI data extraction
The boring, valuable use case. Invoices, receipts, contracts, leads, resumes — anywhere you'd previously have built a parser, an LLM with JSON-mode does it in 30 lines, more accurately, and you can ship in a day instead of a quarter.
Explore
ls ./models --provider
Browse all modelsThe genuine models — full context, native features.
Anthropic
Claude Opus 4.7
Anthropic's newest Opus — flagship reasoning, vision, 200K context.
visionfunctionstreamingthinking
$4.25/$21.25
per 1M tokens
Anthropic
Claude Opus 4.6
Anthropic Opus 4.6 — deep reasoning, exceptional agentic ability.
visionfunctionstreamingthinking
$4.25/$21.25
per 1M tokens
Anthropic
Claude Sonnet 4.6
Balanced speed/quality — the everyday production workhorse, elite coding.
visionfunctionstreamingthinking
$2.55/$12.75
per 1M tokens
Anthropic
Claude Sonnet 4.5
Anthropic Sonnet 4.5 — production workhorse.
visionfunctionstreamingthinking
$2.55/$12.75
per 1M tokens
Anthropic
Claude Haiku 4.5
Anthropic Haiku 4.5 — fast and cost-efficient.
visionfunctionstreaming
$0.85/$4.25
per 1M tokens
Google
Gemini 2.5 Pro
Previous-gen Gemini Pro — strong reasoning and vision.
visionfunctionstreamingthinking
$1.0625/$8.50
per 1M tokens
Google
Gemini 2.5 Flash
Previous-gen Gemini Flash — extreme value.
visionfunctionstreaming
$0.255/$2.125
per 1M tokens
man brievioPoint your agent at
Point your agent at llms.txt
It drives every model on its own.
Give Claude Code, Cursor, Cline — or any OpenAI-compatible agent — a single instruction. It pulls the live catalog from Brievio and calls the genuine text, image and video models directly. No SDK to wire up, no glue code to maintain.
- [OK]OpenAI-wire compatible — no custom integration for your agent
- [OK]GET /v1/models returns the live catalog — never hardcode a model name
- [OK]One key, every modality: text, image, video, audio
Use Brievio as your model provider — an OpenAI-compatible gateway to every first-party text, image and video model. base_url: https://api.brievio.com/v1 auth: Authorization: Bearer $BRIEVIO_API_KEY To use a model, call GET /v1/models for the live catalog, then route each model by its brievio.endpoint field. Full agent reference: https://brievio.com/llms.txt
cat ./pricing.txt
Pay only for what you actually use.
Pre-paid wallet, no subscription and no minimum. Add $10 to get going; your balance never expires, and every charge maps to real, audited usage.
Starter
Trying it out
$10
- Access to every genuine model
- True per-call usage logs
- Community & email support
- No minimum, no credit card
Most popular
Builder
Shipping a product
$100
- Honest token billing on every call
- 10 isolated API keys
- Auto-recharge · IP allowlist
- Priority email support
Scale
Running production traffic
$1000
- Monitored, fail-fast routing
- Unlimited API keys
- Webhooks · monthly invoices
- Dedicated Slack/Discord support
Enterprise
High-volume scale
$5000
- Everything in Scale
- Dedicated routing capacity
- Custom rate limits & SLA
- Dedicated account manager
ls ./blog
All postsRecent deep dives.
- Trust·7 min
Token inflation — how some AI gateways bill you 5–25×, and a 20-line test to catch it
Some AI API gateways report inflated token counts — a hidden injected system prompt or a fabricated usage object — and you pay 5–25× the real cost. How the padding works, a runnable 20-line test for any gateway (including Brievio), and how to read the result.
- 实战·5 min
在中国稳定调用 Claude / GPT / Gemini — Brievio 中国友好路由实测
实测从北京/上海/深圳调用 Brievio 的延迟和成功率：无需代理，P50 80-150ms，成功率 99.9%+。调用的是货真价实的一方模型，按真实 token 计费。支付宝/微信支付/双币卡都可用。健壮重试代码 + 何时真的需要代理。
- 教學·6 min
用 Veo 3 與 Sora 為台灣品牌做短影片廣告 — 5 分鐘完整教學
從文字 prompt 到 9:16 直式 Reels / TikTok / IG 短影片，全流程教學。圖生影片用既有產品照片動起來。5 種台灣品牌實際應用、繁中文字渲染注意事項、商業可用品質的進階技巧。
brievio --helpThe questions
The questions
worth asking.
Didn’t find your answer? Email us at contact@brievio.com — we reply within 24 hours.
Yes — genuine first-party models, not template proxies dressed up to look like them. You get the full context window and the native features: tools, vision and prompt caching. The model that answers your request is the same model the provider ships.
One key, one bill and one OpenAI-compatible endpoint instead of three contracts, three SDKs and three dashboards — plus a small discount on top. Switch models by changing a single string, and you never have to integrate a new client to reach a new provider.
A modest margin on volume infrastructure lets us list at roughly 15% under each provider’s rate — and top-up bonuses take the effective discount to about 21% — not 80% under. When a gateway claims 80% off, ask where the capacity comes from: that pricing usually means gray-market supply that disappears without warning. We’d rather stay up.
Yes. max_tokens and the usual limits are honored natively, exactly as the provider defines them — they aren’t silently ignored or rewritten on the way through.
Every request is logged with the true input and output token counts read straight from the model, never padded by hidden system prompts. Failed requests aren’t billed, streaming responses that drop mid-flight bill only for tokens actually delivered, and you can export the whole ledger as CSV.
No. Your requests and responses are not used to train any model. They’re retained only for your own usage logs and debugging, and they’re scoped to your account.
Yes — it’s a drop-in for the OpenAI SDK. We implement /v1/chat/completions, /v1/embeddings, /v1/images/generations, /v1/models and /v1/video/generations, with streaming, function calling, tool use and vision all behaving identically. Point base_url at https://api.brievio.com/v1 and you’re done.
$ brievio init --production
Build on models that stay up.
Create a key and point the OpenAI SDK at Brievio — the genuine Claude, Gemini and top image/video models, on infrastructure that holds, billed on real tokens. One base URL change and you’re live.