Novita AI Adds Hy4 Flash With Vision at One-Sixth the Input Cost

Novita AI has made Hy4 Flash available through its serverless API, offering text and image input, a 977K context window and an input rate of $0.07 per 1M tokens. The company positions the model as a lower-cost option for agentic, long-context and screenshot-heavy workloads.
- Hy4 Flash is available as tencent/hy4-flash on Novita AI.
- Input pricing is $0.07 per 1M tokens, about one-sixth of Hy4’s current rate.
- The model accepts text and images and offers a displayed 977K context window.
- It supports function calling, structured outputs and reasoning.
- Novita recommends testing Flash and escalating difficult tasks to Hy4.
Hy4 Flash pricing and limits
Hy4 Flash is available under the model ID tencent/hy4-flash through Novita AI’s OpenAI-compatible serverless API. Novita lists pricing of $0.07 per 1M input tokens, $0.281 per 1M output tokens and $0.003 per 1M cached reads.
The model page lists a 1,000,000-token context window and a 64,000-token maximum output, while displaying 977K and 63K respectively. Novita advises users to plan against the displayed figures. The limits match the figures listed for Hy4 and Hy4 Preview.
At the current listed rate, Hy4 Flash’s input cost is about one-sixth of Hy4’s $0.417 per 1M input tokens. Its output rate is about one-quarter of Hy4’s $1.2505 rate. Hy4 Preview is listed at $0.834 for input and $2.501 for output per 1M tokens.
Image input and developer access
Hy4 Flash accepts text and images and produces text. Novita lists chat/completions, responses and Anthropic-compatible endpoints, along with function calling, structured outputs and reasoning support. The service uses the base URL https://api.novita.ai/openai, and developers set the model field to tencent/hy4-flash.
Novita says Hy4 Flash, Hy4 and Hy4 Preview are separate serverless models, so applications must use the intended model ID. Hy4 and Hy4 Flash accept images, while Hy4 Preview is text-only. The company has not published a parameter count for Flash.
For image requests, the source says developers can send images through standard chat completions content parts using an image URL or a base64 payload. Novita recommends testing representative screenshots or other images before routing larger volumes.
Where Novita AI says Flash fits
Novita describes Hy4 Flash as a fit for input-heavy long-context work, including repository analysis, document review and multi-source summarization. Its image support also targets screenshot, diagram and document-page workflows.
Function calling and structured outputs make the model suitable for agent loops, where an application validates tool arguments. The listing also includes reasoning support, although it does not specify reasoning modes or budgets.
The company recommends keeping Hy4 for tasks that require its strongest reasoning. It also cautions that output-heavy requests will see savings closer to one-quarter of Hy4’s output rate, rather than the one-sixth input comparison. Developers can route routine steps to Flash and escalate difficult tasks to Hy4, while tracking which model handled each request.
Hy4 Flash has been live since November 16, 2026. Its lower price and vision support give developers a cheaper way to test long-context and image-assisted workloads, but the source says users should compare task success, tool calls, retries and cost per task before treating it as a replacement for Hy4.
Novita AI: Hy4 Flash on Novita AI: Vision Support at 1/6 the Cost of Hy4


