---
type: "post"
id: "iris-3b-is-a-3b-pixel-space-text-71c4cd"
slug: "iris-3b-is-a-3b-pixel-space-text-71c4cd"
title: "Iris-3B is a 3B pixel-space text-to-image diffusion transformer released under Apache 2.0 by Speridlabs."
description: "**Thesis:** Iris-3B makes a pixel-space generative prior available as an open model that also supports fine-tuned dense-vision tasks."
retrieval_nugget: "**Thesis:** Iris-3B makes a pixel-space generative prior available as an open model that also supports fine-tuned dense-vision tasks."
published_at: "2026-10-12"
updated_at: "2026-10-12"
record_date: "2026-10-12"
date_kind: "published_at"
topics: []
source_urls: ["https://huggingface.co/speridlabs/iris-3b"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:iris-3b-is-a-3b-pixel-space-text-71c4cd"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:ed9e2ffa91d390daca65a64d0067b047aa1cf18e9d46b132feb8e84cfb85a449","reviewed_at":"2026-10-12T20:05:19.856Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":2,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Iris-3B makes a pixel-space generative prior available as an open model that also supports fine-tuned dense-vision tasks.","observed_facts":[{"text":"The Hugging Face model card identifies Iris-3B as a 3-billion-parameter model that outputs pixels directly, without a VAE or latent space.","source_urls":["https://huggingface.co/speridlabs/iris-3b"]},{"text":"The card lists fine-tuned versions for monocular depth estimation, image restoration, and upscaling.","source_urls":["https://huggingface.co/speridlabs/iris-3b"]},{"text":"The card describes Iris-3B Turbo as a four-step distilled version and reports about 0.7 seconds for a 1024×1024 image on one H100.","source_urls":["https://huggingface.co/speridlabs/iris-3b"]}],"mechanism":"Rather than compressing an image into a latent representation and relying on a separate decoder, the model card says Iris-3B predicts pixels directly. The team presents that generative prior as reusable: fine-tuning it for dense tasks is described as requiring no architectural change. The Turbo variant trades the base sampling process for a distilled four-step schedule.","why_now":"Public weights and code under Apache 2.0 make the approach inspectable and testable beyond a hosted demo. The release also supplies an unusually concrete implementation path, while its GPU and storage requirements define the practical cost of trying it.","implications":["Builders can test whether a pixel-space generative backbone is useful across generation and dense vision in their own pipelines.","The reported Turbo speed is a useful hardware-specific reference, not a general latency guarantee; the model card alone does not establish superiority over specialist vision systems."],"evidence_boundary":"The claims and speed figure here come from the project’s Hugging Face model card. They are author descriptions and reported benchmarks, not independent replication; the source excerpt does not establish comparative downstream performance across deployment settings.","watch_conditions":["Whether independent evaluations reproduce the reported Turbo latency and show that the shared pixel-space prior is competitive on depth estimation and restoration under realistic hardware and memory budgets."],"related_records":[],"new_branch_reason":"No prior New Runtime record is supplied for this specific release; this entry begins a distinct record on an open pixel-space model spanning image generation and fine-tuned dense-vision tasks."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/iris-3b-is-a-3b-pixel-space-text-71c4cd/","markdown":"https://newruntime.com/posts/iris-3b-is-a-3b-pixel-space-text-71c4cd.md","json":"https://newruntime.com/posts/iris-3b-is-a-3b-pixel-space-text-71c4cd.json"}
---

# Iris-3B is a 3B pixel-space text-to-image diffusion transformer released under Apache 2.0 by Speridlabs.

## Retrieval answer

**Thesis:** Iris-3B makes a pixel-space generative prior available as an open model that also supports fine-tuned dense-vision tasks.

**Thesis:** Iris-3B makes a pixel-space generative prior available as an open model that also supports fine-tuned dense-vision tasks.

**Observed facts:** The Hugging Face model card identifies Iris-3B as a 3-billion-parameter model that outputs pixels directly, without a VAE or latent space. It lists fine-tuned versions for monocular depth estimation, image restoration, and upscaling. The card says text-to-image weights are about 12 GB and inference requires an NVIDIA GPU with CUDA. It also describes Iris-3B Turbo as a four-step distilled version and reports about 0.7 seconds for a 1024×1024 image on one H100.

**Mechanism:** Rather than compressing an image into a latent representation and relying on a separate decoder, the model card says Iris-3B predicts pixels directly. The team presents that generative prior as reusable: fine-tuning it for dense tasks is described as requiring no architectural change. The Turbo variant trades the base sampling process for a distilled four-step schedule.

**Why now:** Public weights and code under Apache 2.0 make the approach inspectable and testable beyond a hosted demo. The release also supplies an unusually concrete implementation path, while its GPU and storage requirements define the practical cost of trying it.

**Implications:** Builders can test whether a pixel-space generative backbone is useful across generation and dense vision in their own pipelines. The reported Turbo speed is a useful hardware-specific reference, not a general latency guarantee; the model card alone does not establish superiority over specialist vision systems.

**Evidence boundary:** The claims and speed figure here come from the project’s Hugging Face model card. They are author descriptions and reported benchmarks, not independent replication; the source excerpt does not establish comparative downstream performance across deployment settings.

**Watch:** Whether independent evaluations reproduce the reported Turbo latency and show that the shared pixel-space prior is competitive on depth estimation and restoration under realistic hardware and memory budgets.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)