Taifoon

OpenAI-compatible chat completion on 71 models: name one in `model` (llama-3.1-8b-instruct, l3-8b-lunaris-v1, mistral-small-24b-instruct-2501, gemma-3-4b-it, qwen2.5-7b-instruct, gpt-oss-20b, phi-4, gemma-3-12b-it, qwen3.5-9b, gemma-3-27b-it, gpt-oss-120b and 59 more) and the 402 prices that model per request; served by DeepInfra, Venice AI, NEAR AI Cloud and Taifoon's own GPU (nvidia/NVIDIA-Nemot

OpenAI-compatible chat completion on deepseek-ai/DeepSeek-V4-Flash, served by DeepInfra (serverless inference) and sold by Taifoon at the provider's price plus a 20 % reseller cut: 0.108 / 0.216 USD per 1M tokens in/out. This request is priced at its prompt and max_tokens. The answer is checked before you are charged.

OpenAI-compatible chat completion on 71 models: name one in `model` (llama-3.1-8b-instruct, l3-8b-lunaris-v1, mistral-small-24b-instruct-2501, gemma-3-4b-it, qwen2.5-7b-instruct, gpt-oss-20b, phi-4, gemma-3-12b-it, qwen3.5-9b, gemma-3-27b-it and 60 more) and the 402 prices that model per request; served by DeepInfra, Venice AI, NEAR AI Cloud and Taifoon's own GPU (nvidia/NVIDIA-Nemotron-3-Nano-4B-

OpenAI-compatible chat completion on Qwen/Qwen3-235B-A22B-Instruct-2507, served by DeepInfra (serverless inference) and sold by Taifoon at the provider's price plus a 20 % reseller cut: 0.108 / 0.66 USD per 1M tokens in/out. This request is priced at its prompt and max_tokens. The answer is checked before you are charged.

OpenAI-compatible chat completion on openai/gpt-oss-120b, served by DeepInfra (serverless inference) and sold by Taifoon at the provider's price plus a 20 % reseller cut: 0.0444 / 0.204 USD per 1M tokens in/out. This request is priced at its prompt and max_tokens. The answer is checked before you are charged.

OpenAI-compatible chat completion on meta-llama/Llama-3.3-70B-Instruct, served by DeepInfra (serverless inference) and sold by Taifoon at the provider's price plus a 3 % reseller cut: 0.103 / 0.3296 USD per 1M tokens in/out. This request is priced at its prompt and max_tokens. The answer is checked before you are charged.

Latest USD price of BTC, ETH, SOL, NQ, ES, YM from the Taifoon venue tape: the close of the last closed 1-minute bar, its time, age and sha256 digest, replayable with one GET. Only symbols the tape holds; a price older than 180 s is never served as current (nothing is charged). BTC, ETH, SOL are held to Chainlink on Base within 1 % when its feed is under 1 h old. 0.003 USDC a call.

OHLCV bars of BTC, ETH, SOL or the CME micro futures NQ, ES, YM (15s, 1m, 5m, 1h) from the Taifoon venue tape, past windows only (up to 1,000 bars a call), with a sha256 digest of the bars. The tape seals a past window, so the same request gives the same bars. The same source is also served over MCP by Cubicle candles (ERC-8004 8453:96559).

A trader’s realized record recomputed by code from its fills (Taifoon venue by attribution, paper; or Hyperliquid by address): fills, notional, PnL before fees, fees, wins, losses, max drawdown, the same round trips held long, and the fills with their sha256 digest. Facts about past fills, not advice. The same source is also served over MCP by Cubicle copy-trading (ERC-8004 8453:96560).

Olas-style yes/no forecast: { question } in, { p_yes, p_no, confidence, info_utility } out with reasoning and sources. BTC/ETH/SOL price questions with a threshold and a date are modelled from the Taifoon venue tape (sources replayable); any other question is a small open model's prior with no web search (no sources, confidence at most 0.3). Checked before you are charged. 0.01 USDC a forecast.

Rules compiled into typed questions inside Taifoon's Intel TDX enclave (Phala Cloud). The answer carries two TDX quotes, both verified by Phala's verifier before you are charged: one binds the result and the enclave key, one binds that key to sha256 of your nonce. The compose-hash in RTMR3 proves which code runs. 0.001 USDC a request. Send { rules, nonce? }.

On-chain data query on Base, sold by Taifoon: the tokens and fee tier of a Uniswap V3 pool, or the hash of a block. An admitted data seller answers; Taifoon reads the same facts from the chain and grades the answer before you are charged. A failed answer is not charged. Answer + graded receipt. 0.03 USDC a call.