Compare the live per-model Flowie rates for text agents, image, video, audio, 3D, rendering, and hosting.
Where: Account → Get Flowies / Plan & Billing → See full Flowie pricing, or /pricing
Key ideas
- One source of truth: The public rows are derived from the same pricing tables used to estimate, gate, charge, and refund runs.
- Correct units: Video is generally per second, images per output, agent chat per turn/action, and other media by their displayed generation unit.
- Model-aware chat: Language-model rows show the Flowie cost of each selectable model instead of quoting one generic chat price.
- Resolution and duration determine current video cost: Seedance 2.5 charges 14/30/72 Sparks per second at 480p/720p/1080p. PixVerse V6 charges 3/4/5/9 per second at 360p/540p/720p/1080p. The live node quote multiplies the selected route, resolution, duration, sound/reference options, and output count.
Steps
- Open the full Flowie pricing dialog.
- Choose a tab such as Text Agent, Image, Video, Audio, 3D, Hosting, or Render.
- Compare the model and resolution tiers.
- Return to the relevant model picker and confirm the same cost badge before running.
- Before a large generation or agent run, check the “from” quote and your balance; after the run, compare the number of rounds and measured input/output rather than expecting the picker floor to be the final total.
Read the unit first
- Language-model/agent rows: per displayed turn, action, or tier.
- Images: per output, sometimes affected by resolution, megapixels, references, or output count.
- Video: usually per generated second, with resolution/model/action variants and possible reference surcharges.
- Audio/music/voice: per displayed duration/output unit.
- Research/analysis: collection or analyzed-row/action units.
- 3D, render, hosting, and utilities: the unit shown in their own row.
The final estimate can vary
Model choice alone may not determine cost. Resolution, duration, quality, start/end/reference inputs, output count, context size, generation versus extend/edit, and provider-specific options can change the Flowie estimate.
Gating and refunds
The same canonical pricing ids feed the pre-run balance gate and charge path. Failed/cancelled operations follow their action’s refund policy; verify the resulting Usage balance rather than assuming every interruption is charged or free.
Read model prices as a floor, not a whole-task promise
Model pickers now say “from” per call. The base round covers an included input/output allowance; long uncached input and measured output can increase the charge. Reasoning is included in measured output and is not multiplied a second time in the measured surcharge. Several agent rounds are billed separately.
- GLM 5.3 Flash has a 1-Spark base round floor; the selected effort and actual usage still matter.
- Cached prompt tokens are excluded from the uncached-input calculation.
- Wan 3.0 rates are 3/6/11 Sparks per output second at 480P/720P/1080P; Prime is 4/8/16. Source video is additionally billed at the selected rate.
- Omni 1.1 rates are 2/6/9/18 per output second at 360p/720p/1080p/4K, plus one Spark per source-video second and one flat Spark for a reference-image run.
- GPT Image 2 is flat-priced across supported sizes: 20 generation / 22 edit.
Current language models and measured pricing
The shared catalog adds Gemini 3.8 Flash, Qwen3.8 Flash, GPT-6 Astra, Claude Haiku 4.5, Claude Sonnet 5, and Claude Opus 5. Saved Gemini 3.5/3.6/3.7 Flash choices now resolve to 3.8. Historical labels remain unchanged, and the visible base round may increase with effort, measured usage, or Astra’s large-context tier.
- Qwen image/video capability is catalogued, but live vision was not verified during the refresh because the upstream route was unavailable.
- Astra/Claude default to OpenRouter for reliable prefix caching; optional Kie-first routing is a server setting.
- The current action quote is authoritative for a paid run.
GPT Image 2.5 quote
- Flare and Sunburst cost 8 Sparks at 1K, 9 at 2K, and 14 at 4K.
- The first edit reference is included; each additional reference adds 1 Spark, up to sixteen total.
Tips
- Make a short or low-resolution representative run before a long final.
- Compare models on required capabilities as well as Flowies; a cheaper model that lacks the needed input can cost more in retries.
Limitations and important notes
- Catalog rates can change as providers and routing change; do not hard-code a help screenshot as a permanent price.
- Some compound agent workflows can invoke separately priced sub-actions.
Related guides
- Flowmo pricing and Flowies: the overview
- Track Flowies, limits, and Powerups
- Choose the language model in any chat
- Flowgen image, video & text model catalog
- GLM 5.3 Flash: multimodal chat, reasoning, and pricing
- Wan 3.0 and Prime: video modes, references, and controls
- Gemini Omni 1.1 Flash: generate, reference, edit, and extend video
- September 2026 language-model refresh: Gemini 3.8, Qwen 3.8, GPT-6 Astra, and Claude 5
- Generate and edit images with GPT Image 2.5 Flare and Sunburst