Models and Pricing

Pricing Page

Model Pricing displays a build-time snapshot and attempts to refresh live prices in the browser.

Models, pricing, and development documentation are unified from the backend model catalog. Models with only catalog metadata, not yet open for calls, or not yet priced are also displayed, clearly marked as “Not available” or “Not yet priced.”

Model Landing Pages

Each model has a dedicated URL, such as /models/deepseek-v4-flash/, including:

• Tier comparison and input/output pricing
• Provider and capability tags
• Copyable call examples

Model Development Documentation

The sidebar “Model Docs” provides development guides (model ID, parameters, examples, availability status, and pricing) for each model in the backend catalog, such as /docs/models/deepseek-v4-flash/.

Adding Generative Models

The current model catalog is categorized by invocation method:

• Text chat:claude-sonnet-5, deepseek-v4-flash, deepseek-v4-pro, gemini-3.5-flash, gemini-3.6-flash, gemini-3.7-flash, glm-5.2, gpt-5.5, gpt-5.6-sol, gpt-5.6-terra, grok-4.5, grok-4.6, minimax-m3
• Image generation: gpt-image-2, grok-imagine-image-2.0, midjourney, nano-banana-2, nano-banana-pro
• Video generation:veo-3.1, seedance-2.0, seedance-2.0-pro, seedance-2.0-fast, seedance-2.0-mini, seedance-2.5, hailuo-3, grok-imagine-video-1.5, gemini-omni, kling-3.0-turbo, kling-o3, happy-horse-1.1, wan-2.7, sora-2, sora-2-pro
• Music generation:suno-v5.5
• Not yet available:nano-banana(catalog entry only, not currently available)

Text models use the OpenAI-compatible POST /v1/chat/completions, with streaming support by default. Music, image, and video tasks are submitted to POST /v1/audios/tasks, POST /v1/images/tasks, POST /v1/videos/tasks, and results are queried via GET /v1/tasks/{task_id} Full request parameters are detailed in Audio Generation Tasks and Video Generation Tasks; actual available models are subject to console channel configuration and /v1/models responses.

The suno-v5.5 in the current catalog is displayed as Suno 5.5 and billed at a fixed backend price;gpt-image-2 is billed per Token, with separate multipliers for text input, image input, and image output. Prices modified by administrators in the console take precedence over code defaults.

Fixed Price Snapshot: 2026-08-21

This batch of EvoLink models is fixed at a RokoAPI selling price based on the provider’s daily USD cost plus 10%, independent of real-time credits at request time. Text models are settled based on actual input, output, and cached Tokens;MiniMax M3 and Grok 4.5/4.6 also select corresponding price tiers based on context length.

Veo 3.1 is billed based on the combination of output count, duration, resolution, and audio toggle for generated videos:

Veo 3.1 supports 4, 6, and 8-second outputs,n supports 1–4; if the request does not pass generate_audio , the upstream preview model generates audio by default, and billing is applied at the “With Audio” tier. Current fixed expressions and price tiers for other text and generative models are subject to Model Pricing.

Model IDs, Series, and Aliases

• API requests use model IDs, such as suno-v5.5.
• Pages are aggregated by series; for example, DeepSeek V4 Flash and Pro share a single series page.
• Midjourney uses the single public model ID midjourney; the legacy IDs Midjourney v7, Midjourney v8.1, mj-v7, and mj-v8.1 are no longer available.
• midjourney offers Fast and Turbo, with Fast as the default. Standard generation costs $0.055 in Fast or $0.077 in Turbo; HD is billed at 1.5× the selected speed price, then reconciled against the upstream quota after completion.
• Suno model IDs must include a version number. The public name for suno-v5.5 is Suno 5.5; the unversioned suno no longer represents the current version.

Combined Parameter Billing

RokoAPI’s unified billing rules account for text tokens, image count, resolution, quality, video output duration, reference material duration, generation mode, and built-in tool call counts. As a result, the same model can incur different costs across combinations such as 1K / 2K / 4K, 480p / 720p / 1080p, text-to-image / image-to-image / reference-to-image, without requiring separate model entries for each capability.

Asynchronous generation tasks pre-charge estimated costs based on fixed price lists and request parameters at submission, then settle the difference based on actual tokens or output duration upon completion. EvoLink video models use a backend-saved price snapshot that does not fluctuate with real-time provider credits; price list updates are published after admin review. Failed tasks are handled according to the task refund policy. For current rules, see Model Pricing.

Notes

• Snapshot content remains readable with JavaScript disabled
• Failed live refreshes do not clear the page; they retain the snapshot and display a prompt
• Price changes can trigger a Pages rebuild via Deploy Hook