Introducing bInference.
AI credits backed by BSC revenue.
Plug bInference Router into a Binance Agent OS agent or into Claude Code, Codex, Muse, OpenClaw or Hermes.
Use them for trading, Agent OS features or R&D.
Claude Fable 5.2 and Gemini 4 Carbon, now on bInference Router.
Fable 5.2: Anthropic's most capable model. Top score in agentic coding.
Gemini 4 Carbon: Google's frontier model. Finds security bugs and patches them.
Both with 1M context, 20% under cheapest market prices.
Two new models on bInference Router, AI Chat and Video Studio.
Tencent Hy4: its new flagship, now with vision and computer use.
Gemini Omni 1.2 Flash: Google's newest video model, 15-second clips.
Use them on any plan, or pay as you go with credits.
bInference Agent is 90% built.
29 days of open-source development.
1,759 commits. 1,559 test files.
280K lines of code.
15 plugins, 25 integrations, one agent for BNB Smart Chain: trades, launches, earns and researches inside your limits.
bRoutes are live on the bInference Router 🔀
Stop paying top-model prices for easy prompts.
Easy ones go to a cheap model, hard ones still get your best.
Cut costs for your chatbots, coding tools and AI agents: easy prompts run up to 40x cheaper per token.
Make your own bRoute:
1. Describe each lane in plain words
2. Pick a model for each
3. Try a prompt to see its lane and cost
4. Call it as a model: broute/<name>
Or describe your app in one sentence and the lanes are drafted for you.
https://t.co/NZJ9tWsada
bRoutes are live on the bInference Router 🔀
Stop paying top-model prices for easy prompts.
Easy ones go to a cheap model, hard ones still get your best.
Cut costs for your chatbots, coding tools and AI agents: easy prompts run up to 40x cheaper per token.
bInference Router just crossed 50 billion tokens.
That's about 37.5 billion words.
7× all of English Wikipedia.
34,000× the whole Harry Potter series.
Nearly 300 years of nonstop reading.
From 4.2 billion to 50 billion in 15 days. Now 2 billion a day.
Every agent on bInference now has its own ID on BNB Smart Chain.
Launch a token, get a running AI agent, and its Agent ID lands in your wallet within minutes.
IDs are issued in BNB Chain's ERC-8004 Identity Registry, the open standard for agent identity.
Gas on us.
GPT-6.1 Terra is live on bInference Router.
On par with GPT-6 Astra at a tenth of the price, and 3x faster.
One Agents' Last Exam task: $8.82 on Claude Opus 5.5, $0.81 on Terra.
Terra and Terra Pro, from $0.80 in and $4.00 out per 1M tokens.
bInference Router has 4 ways in.
AI Chat on monthly plans. Clips, Video and Audio studios. API keys for builders. Hosted Agents on BNB Chain. All powered by the bInference Router.
200+ users in 30 days.
$20K+ of AI served.
30 Billion+ tokens routed.
Kimi K3.1 Pro is live on bInference.
In Chat on every plan and on the bInference Router.
#1 on SWE Marathon. #2 on Terminal-Bench 4.0
0.7 behind Claude Opus 5.5, at about a third of the cost per task.
One of the first to serve it. All private.
Recap: bInference, less than a month live.
16 launches. 152 users. Almost 30 Billion tokens routed.
Router, agents, private AI, Clips, Audio, Inference DEX, free models and more. All live.
Next: bGPU.
Introducing bInference ZeroRoute: free AI models, one endpoint.
Call "binference/free" and Capability-Matched Rotation picks a free model that can handle your request: images, tools or JSON.
17 free models from NVIDIA, Google, MiniMax and more. $0, a new key works.
bGPU waitlist is open.
bInference GPU market, anonymous by design. No account, no KYC, just your BSC wallet.
Rent RTX 4090 to H100, to train models, run your own LLMs or generate images and video.
Host your idle GPU, get paid daily in USDT.
DeepSeek V4.1 Flash is live on the bInference Router, from day one.
Served by Together, a US-based provider. Same model, US infrastructure.
Beats Opus 5 on DeepSWE (77.0 vs 74.0). 16× cheaper than Opus 5.5.
Connect your BSC wallet and use it in Private mode.