On 12 May 2026, Krea announced Krea 2, including K2 Raw and K2 Turbo. Krea prioritised controllable aesthetic range and later exposed both its raw foundation checkpoint and accelerated production variant.
This release brief was checked against first-party material on 10 August 2026. The date above is the public announcement date, not the date a repository was created or a third-party provider added the model. Where access or weights arrived later, that distinction is recorded below.
Release record
| Field | Verified detail |
|---|---|
| Announcement | 12 May 2026 |
| Availability or weight release | 12 May 2026 in Krea; permissively licensed weights and report on 23 June 2026 |
| Release type | hosted and open-weight text-to-image foundation-model family |
| Access | Krea product access plus downloadable K2 Raw and K2 Turbo weights, inference code and training report |
| Architecture | a roughly 12B single-stream multimodal diffusion transformer with a rectified-flow objective and a fast Turbo variant |
| Maximum stated context | image-generation inputs and style references rather than a published text-token context limit |
What changed
Krea 2 matters because the company released more than a polished image endpoint. Its technical report documents data curation, transformer ablations, autoencoder choices, distributed training and the distillation path for Turbo, while the weights make independent creative-control testing possible. The product also treats style references as adjustable inputs that can be combined and weighted, shifting evaluation from generic prompt adherence toward repeatable art direction.
The practical comparison is therefore not simply whether Krea 2, including K2 Raw and K2 Turbo has the largest headline score. Teams need to ask whether its architecture, access terms, latency, tool behaviour and evaluation setup match the workload they actually intend to run. A model can lead one harness while losing on cost, refusal behaviour, multilingual quality or repeatability in another.
Benchmarks worth retaining
| Evaluation | Reported result | How to read it |
|---|---|---|
| Model scale | about 12B parameters for the foundation model | Capacity figure from Krea’s release materials |
| Turbo latency | about 2 seconds in the hosted Krea 2 Turbo launch | Product latency claim that depends on serving conditions |
| Resolution training | evaluated through 256px, 512px and 1024px stages | Training and ablation envelope, not a preference score |
These are release-time results, not independently reproduced guarantees. Visual preference is sensitive to prompt set, sampler, resolution and rater population, and the May hosted model should not be assumed identical to every checkpoint published in June. Scores should remain attached to the disclosed effort setting, agent harness, tool access, timeout, context-management policy and judge model. Moving a number into a procurement sheet without those conditions creates false comparability.
Architecture and access
Krea 2, including K2 Raw and K2 Turbo is described as a roughly 12B single-stream multimodal diffusion transformer with a rectified-flow objective and a fast Turbo variant with image-generation inputs and style references rather than a published text-token context limit of stated context. Its access position at verification time is Krea product access plus downloadable K2 Raw and K2 Turbo weights, inference code and training report. That wording matters: open weights, source-available weights, an API, a product preview and a research demonstration give adopters very different rights and different levels of reproducibility.
Before deployment, record the exact model identifier or checkpoint, inference stack, quantisation, reasoning setting, region, price schedule and supplier terms. If the release uses a custom licence, read the licence itself rather than relying on the word “open” in launch copy. If it is API-only, preserve the dated documentation and change-notice route because the served snapshot can change without a downloadable artefact.
What an evaluation should test next
For Krea 2, including K2 Raw and K2 Turbo, a credible internal gate should include:
- a frozen set of representative tasks with pass, fail and abstain criteria;
- a matched baseline using the same tools, timeout, prompt budget and reviewer rubric;
- repeated runs to expose variance rather than reporting a single best attempt;
- latency, token use and total task cost alongside task success;
- adversarial, multilingual and long-context cases relevant to the real deployment; and
- rollback evidence showing the previous model can be restored safely.
The wider model change-control guide explains how to keep model, prompt, tool and corpus changes reconstructable. The AI dependency inventory guide covers the release and supplier records needed after deployment.
AIEngine verdict
Krea 2 qualifies as a major open generative-model release. Creative teams should evaluate style controllability and subject consistency with saved seeds and exact checkpoints, not a mood-board comparison of hand-picked outputs.
This is a launch assessment, not a certification. Benchmark leadership is useful evidence of where to test; it is not authorization to place the model in a high-impact workflow without domain evaluation, security review and an accountable owner.
Primary sources
Image provenance
Hero image: Krea official release artwork. The locally served WebP is a crop of the first-party release or model-card asset recorded in the repository provenance manifest.
