Adobe Firefly now supports third-party AI models for image and video generation. Builders gain flexibility but face new model management decisions.

Builders get model flexibility within Adobe's interface without losing the operational simplicity of integrated workflows.
Signal analysis
Adobe Firefly has opened its generative features to non-Adobe models. Previously, Firefly relied exclusively on Adobe's own trained models. Now builders can select from partner models when using Generate images, Generate video, Firefly video editor, and Boards—Adobe's collaborative workspace tool.
This isn't a wholesale replacement of Adobe's models. Instead, it's an opt-in expansion. Users can choose which model powers their generation task based on the specific use case. The integration happens within Adobe's existing UI, meaning no context-switching or API wrapping required.
Model choice adds a decision point. Builders now need a strategy: use Adobe's models for consistency and optimization within the ecosystem, or switch models based on task requirements (speed, style, output format). This is operationally simpler than managing external APIs, but it requires builders to know the tradeoffs between available models.
Pricing likely varies by model. Adobe hasn't publicly detailed pricing tiers for partner models versus native Firefly, but this is a critical variable. Builders need to understand cost-per-generation before scaling any workflow that relies on specific partner models. Switching models mid-project could change unit economics.
This move signals Adobe's shift toward being a platform orchestrator rather than a single-model vendor. For builders already invested in Adobe's ecosystem, this increases stickiness—you get model optionality without leaving the interface. For builders considering Adobe's tools, the expanded model access is a material differentiator versus competitors.
Adobe is responding to model commoditization. Firefly was a differentiator 18 months ago. Now, image and video generation models are table-stakes. By integrating partners, Adobe keeps builders from churning to competitors while maintaining its position as the primary interface.
Vertical integration is breaking down. Adobe can't compete on model quality alone anymore—it needs to aggregate options. This reflects the reality that no single model wins across all use cases. Builders want speed for some tasks, quality for others, specific style consistency for brand work. One model doesn't cover all.
The partnership model creates switching costs. If builders select partner models within Adobe's interface, they become dependent on that integration. Adobe benefits from increased usage lock-in without needing to own or train competitive models.
First, audit your current Firefly usage. Are you using it for speed, quality, consistency, or some combination? This determines whether you need to test partner models or stick with Adobe's defaults. If you're already satisfied with output quality, there's no immediate urgency to experiment.
Second, map model options to your actual workflows. Don't test every partner model available. Instead, identify 2-3 use cases where you suspect a partner model might perform better (faster generation, better style matching, lower latency). Run A/B tests with fixed inputs, then measure output quality and cost per use.
Third, document pricing if Adobe publishes it. Partner models may cost differently than Firefly's standard offering. Build cost assumptions into your project estimates. If a partner model is 20% faster but costs 50% more per generation, that's a trade-off you need to quantify upfront.
Best use cases
Open the scenarios below to see where this shift creates the clearest practical advantage.
One concise email with the releases, workflow changes, and AI dev moves worth paying attention to.
More updates in the same lane.
Discover how to enable Basic and Enhanced Branded Calling through Twilio Console to enhance your brand's visibility.
Cohere has unveiled 'Cohere Transcribe', an open-source transcription model that enhances AI speech recognition accuracy.
Mistral AI has released Voxtral TTS, an open-source text-to-speech model, providing developers with free access to its capabilities for various applications.