Phi 4
Neural Network
Phi 4 is a compact 14B parameter model from Microsoft Research, a Phi 4 API for complex reasoning tasks.
Max answer length
(in tokens)
Context size
(in tokens)
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Phi 4?
Phi 4 is a 14-billion parameter language model from Microsoft Research designed for complex reasoning. It handles multi-step tasks, maintains logical consistency, and remains compact, making it reliable where memory is limited or response speed is critical. The focus is on reasoning quality rather than size. In BotHub, the model is accessible without a VPN or foreign card: pay with a Russian card in rubles based on actual token usage, not a subscription, and your balance never expires. With over 250 models in one window, you can easily compare Phi 4's output with others and switch instantly. The unified OpenAI-compatible API lets you change models without rewriting your integration. Chats are encrypted, and corporate clients get contracts, invoices, EDI, and an admin panel with limits. Phi 4 helps analysts untangle multi-step tasks, engineers verify algorithm logic, teachers break down solutions, and product teams quickly test hypotheses where speed and cost matter.Frequently asked questions about Phi 4
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
Phi-4 is a compact text model from Microsoft. It accepts not only text but also images and documents, making it suitable for analyzing files, extracting data from diagrams and scans, writing and editing text, coding, and solving logical and mathematical problems. Available from Russia without a VPN or foreign card.
The maximum response length is 14,745 tokens, enough for a long article, documentation, or detailed analysis. Keep in mind the total context window of 16,384 tokens, which includes your request, attachments, and the response itself. It is better to break large tasks into parts.
A dedicated step-by-step reasoning mode is not explicitly noted in our data, but that doesn't mean the model can't handle logic. You can usually get a detailed breakdown directly in the prompt: ask it to break down the solution into steps and show intermediate conclusions, not just the final result.
There is no explicit note about support for function calling or structured JSON output in our data, so we cannot guarantee it. You can request a JSON response directly in the prompt. If you need a strict schema and tools, switch to another model in the unified BotHub API without rewriting your integration.
You can upload images and documents: the model will analyze a scan, diagram, table, or PDF and respond with text. Audio and video are not listed as supported inputs, but BotHub has separate models for them—you can switch to them in the same window.
We don't have specific data on Russian language quality, so we won't make promises in advance. The most reliable way is to send your typical request and see the result. If the phrasing seems weaker than expected, compare the response with another model in the same BotHub window without changing your integration.