Shieldstral 1.0 (3B)
by Mistral AI
Shieldstral 1.0 (3B) is a model from Mistral AI, released on August 4, 2026. It ranks #200 of 220 models by Intelligence Score with a score of -2.4.
Shieldstral is a 3B open-weight multimodal safety classifier from Mistral. It frames content moderation as policy-adaptive yes/no question answering: plain-language policies are supplied at inference time, and the model returns a calibrated safety score from a single forward pass for text, image, or text+image content. Released under Apache 2.0. Aggregate self-reported average F1: text-safety 0.849, multimodal 0.838 (HF mistralai/Shieldstral-1.0-3B / arXiv 2607.25857); per-bench WildGuard/HarmBench/ToxicChat/BeaverTails/Aegis/VLGuard ids are not in the catalog.
Key specs
| Provider | Mistral AI |
|---|---|
| Release date | 2026-08-04 |
| License | open source |
| Intelligence Score | -2.4 |
Quick answers
Is Shieldstral 1.0 (3B) open source?
Yes. Shieldstral 1.0 (3B) is an open model released under the open source license.
When was Shieldstral 1.0 (3B) released?
Mistral AI released Shieldstral 1.0 (3B) on August 4, 2026.
How does Shieldstral 1.0 (3B) compare to Claude Opus 5.5 (Reasoning)?
Shieldstral 1.0 (3B) scores -2.4 on the Intelligence Score versus 60.5 for Claude Opus 5.5 (Reasoning), the current #1.
Compare on the AI leaderboard · All AI models
More models from Mistral AI
- Mistral Large 4 — Mistral AI
- Mistral Medium 3.5 (Reasoning) — Mistral AI
- Ministral 3 (14B Reasoning 2512) (Reasoning) — Mistral AI
- Mistral Small 4 (Reasoning) — Mistral AI
- Ministral 3 (8B Reasoning 2512) (Reasoning) — Mistral AI
- Min istral 3 (3B Reasoning 2512) (Reasoning) — Mistral AI