Kiyotaka.ai unveils free LLM APIs
- Kiyotaka’s current launch is a free tier for its market-data API, not an LLM API, giving developers programmatic access to crypto and Polymarket feeds. - B.AI’s live product is a multi-model chat and API layer with OpenAI-, Anthropic-, and Google-compatible access plus blockchain-based top-ups and wallet login. - The common thread is cheaper access to agent plumbing, especially model routing and data feeds for builders outside big platforms. (docs.b.ai)
Kiyotaka’s live product is a free API tier for trading and prediction-market data, not a newly documented large language model API. Its docs say API keys are included on Free, Basic, Standard, and Advanced plans. (docs.kiyotaka.ai 1) (docs.kiyotaka.ai 2) The API is built around market data: candles, open interest, funding rates, liquidations, heatmaps, volume profiles, and Polymarket analytics. Kiyotaka says those feeds run through a single endpoint with API-key authentication. (docs.kiyotaka.ai) (kiyotaka.ai) Its quickstart example shows a REST call to `api.kiyotaka.ai/v1/points` for hourly BTC-USDT data from Binance Futures. The docs also list WebSocket streams for real-time liquidations and open interest. (docs.kiyotaka.ai 1) (docs.kiyotaka.ai 2) (docs.kiyotaka.ai 3) Kiyotaka also ships kScript, its own scripting language for indicators and screening. The company pitches it as accessible to first-time coders as well as traders coming from PineScript. (docs.kiyotaka.ai) (kiyotaka.ai) B.AI, by contrast, is actually selling model access. Its documentation describes an “LLM Service” with multi-model chat, API access, wallet-based login, Google sign-in, and blockchain payment flows. (docs.b.ai 1) (docs.b.ai 2) The API is OpenAI-compatible on chat completions and supports either bearer tokens or `x-api-key` authentication. B.AI lists ` as the base URL and says the same secret can be used for both auth methods. (docs.b.ai) Its model catalog includes OpenAI, Anthropic, and Google families alongside Chinese labs such as Zhipu, Moonshot, and MiniMax. The pricing page lists GPT-5.4, Claude Opus 4.6, Claude Sonnet 4.6, Gemini 3.1 Pro, GLM-5, Kimi K2.5, and MiniMax M2.5. (docs.b.ai) (docs.b.ai) (docs.b.ai) B.AI settles usage in a unified credit system instead of billing each model separately in dollars. Its docs say accounts are prepaid and can be topped up on the TRON network with TRX, USDT, USDD, and USD1. (docs.b.ai) (docs.b.ai) That makes the cleaner read on this story less about one company launching “free LLM APIs” and more about two adjacent pieces of agent infrastructure showing up at once. One is a free data pipe for markets; the other is a wallet-linked switchboard for frontier models. (docs.kiyotaka.ai) (docs.b.ai) For builders, the practical change is that data access and model access are getting unbundled into simpler services. You can buy model calls from one vendor, market or event data from another, and stitch the workflow together with standard APIs. (docs.b.ai) (kiyotaka.ai) The result is a lower-friction stack, but the underlying products are not the same thing. Kiyotaka is selling structured market data with a free entry tier; B.AI is selling multi-model inference with crypto-native payments. (docs.kiyotaka.ai) (docs.b.ai)