Alpha
Contact
Implement · Soofi

Can you run it?

Ownership levelPartialnone·limited·partial·substantial·fullAnalytical input C · 63.2/100

This page is a projection of the one entry record, the Reliability and Data control factors that Implement covers. The full verdict is set by all four factors together, floor-weighted so the weakest caps the whole.

Which domain expands which factor
  • AssessUse & modify + Transparency
  • ImplementData control + Reliability
  • UseReliability
  • SupportTransparency

Install & run

Soofi-S ships as bf16 safetensors on the verified Soofi-Project org, plus first-party GGUF builds, but every repo is manually gated. Request access and wait for publisher approval before any file downloads; the card describes a closed beta for selected partners. The custom hybrid Mamba-2/MoE needs trust_remote_code=True on the transformers path. Pin an exact revision.

# Hugging Face transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Soofi-Project/Soofi-S-Instruct-Preview"   # deploy the post-trained variant
rev = "<pinned-commit-sha>"                              # pin the revision, don't float on main
tok = AutoTokenizer.from_pretrained(model_id, revision=rev, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_id, revision=rev, trust_remote_code=True)
# llama.cpp / llama-server (first-party GGUF) - use --jinja for the native tool template
./llama-server -m soofi-s-instruct-preview.Q4_K_M.gguf --jinja

# Ollama (community/first-party GGUF)
ollama run soofi-s-instruct

# vLLM (OpenAI-compatible server) - confirm your vLLM version supports the hybrid arch
vllm serve Soofi-Project/Soofi-S-Instruct-Preview --revision <sha> --trust-remote-code

Prefer safetensors; verify per-file checksums after download.

Hardware & VRAM requirements

The MoE activates only ~3B of 30B parameters per token, so decode is cheap, but the full ~30B weight set must be resident. In bf16 that is roughly 60 GB, so plan for multi-GPU or a first-party GGUF or 3-bit quantization to fit a single high-VRAM card. The hybrid Mamba-2 design keeps the inference cache near-constant as context grows, so serving up to 256k tokens does not blow up VRAM the way a dense transformer's KV cache would. Size for the largest variant to serve and budget headroom.

Serving stacks

Runtime support for the custom architecture is still maturing, so check the stack's version before committing:

StackUse it for
llama.cpp / llama-serverGGUF inference; --jinja applies the embedded chat/tool template verbatim
vLLMHigh-throughput + OpenAI-compatible API (needs --trust-remote-code; confirm hybrid-arch support)
OllamaLocal single-command runs (note: its template omits the native tool format)
transformersPrototyping and fine-tuning (trust_remote_code=True)

Operational Readiness scores 4. The breadth is there, but the custom Mamba-2/MoE and cross-runtime tool-template quirks need more care than a plain Llama-class model.

Safe-deployment controls & Deployment Ceiling

Apply the safe-deployment playbook. The model-specific gate rows are:

  • Supply chain: pin the exact Soofi-Project revision hash and verify checksums; prefer safetensors. trust_remote_code=True runs vendor modelling code, so review it.
  • Model layer: deploy the Instruct-Preview variant, never the base checkpoint, and set conservative decoding defaults.
  • Input/output layers: add guardrails and your own guard classifier for any customer-facing path, because Soofi-S ships no companion guard model and its safety features are documented as incomplete. Add prompt-injection defences and treat retrieved and tool content as untrusted.
  • Compliance: archive the per-source training-content summary and copyright posture in your AI-BOM; the source data is readily available.

Deployment Ceiling: T2 (customer-facing, human-reviewed), conditional and preview-caveated. With the Instruct-Preview variant, input/output guardrails and prompt-injection defences, Soofi-S can in principle reach customer-facing, human-reviewed use. It is an early preview with incomplete safety hardening, so pilot it internally first. Its strengths are European sovereignty, auditability and long-context economics. T3+ (bounded autonomous) also needs a deterministic action-policy engine, per-action human approval for irreversible operations and external guard tooling beyond what this release provides.

Available quantizations

The Soofi-Project org publishes first-party GGUF builds for the preview variants and an EntQuant 3-bit build, so quantized inference does not depend solely on community efforts. Coverage is good for a preview. Test quality against the bf16 base for your workload before relying on an aggressive (e.g. 3-bit) quant in production.

Fine-tuning & adaptation

Soofi-S is built for fine-tuning, but the rights are not there yet. The training code is public (soofi-project/Soofi-Pretraining, with no licence file; soofi-trainer, MIT) and the per-source data accounting documents the recipe, so the project could in principle be reproduced or extended. The weights and the intermediate Megatron checkpoint are manually gated, and no licence in force grants adaptation or redistribution of derivatives. Confirm the terms of your access approval before fine-tuning. With access, standard SFT/LoRA via transformers / TRL on the safetensors checkpoints is the practical path. Placing a derivative on the market carries EU AI Act provider implications (see Assess).

API / OpenAI-compatible integration

Serve Soofi-S behind the OpenAI-compatible endpoint of vLLM or llama-server for drop-in integration with existing OpenAI-client code:

llama-server -m soofi-s-instruct-preview.Q4_K_M.gguf --jinja --port 8000
# then point any OpenAI SDK at http://localhost:8000/v1

Use --jinja (llama.cpp) so the model's embedded chat/tool template is applied verbatim, which is what makes native tool calling work. Some runtimes (e.g. Ollama) strip the tool portion, so verify the template your endpoint applies.

How this scores

The ownership factors this domain covers, drawn from the one entry record.

3

ReliabilityIs it reliable and good enough for the job?

Moderate

Pre-release: the weights are available only through manual approval, so runtime dependability is unproven for most adopters and the documented-incomplete safety cannot yet be exercised. The reported ~3B-active-parameter strength counts as a capability point in the AOI score, not as an ownership factor.

How this scores (AOI sub-dimensions)
Operational4/5how practical it is to run, serve and maintain in productionRuns on mainstream stacks (HF transformers with trust_remote_code, vLLM, llama.cpp/llama-server and Ollama) with first-party GGUF and 3-bit quantizations and a near-constant long-context cache.
Safety3/5whether misuse risks are evaluated and guardrails are providedA post-trained Instruct-Preview variant exists and the L3S Research Centre contributes safety and evaluation frameworks.
4

Doesn't extract your dataDoes running it keep your knowledge and data yours?

Moderate

EU/Germany-domiciled and trained on sovereign German infrastructure, so once self-hosted your data would stay yours. With weights manually gated and no licence in force, that self-hosting cannot yet be exercised from a settled grant.

How this scores
Not a scored AOI dimension. For a self-hosted model, data-control is a structural property of running the weights yourself, strong by default unless the model phones home or the licence claws back rights. For a hosted API this factor is the retention + train-on-inputs + residency read, scored from the binding terms.
What this means for adoptionToday you do not own the use of these weights. They are manually gated in a closed beta, the card says it is not an open release, and no licence is in force, so no published text grants use, modification or commercial use. What exists is thorough documentation: a pretraining report with full per-source data accounting and a released training codebase. A stated OSAID 1.0 commitment has no date. The marketing says open-source; the binding text says closed beta. Treat openness as promised, not delivered, and re-assess when weights and a licence are published.

Sources

The same evidence records as the entry sheet. Read means the text was verified; unverified means it is known to exist but not yet read.

Technical_reportread2026-07-25
The arXiv report ("A Sovereign, Open-Source Foundation Model for German and English"; PDF "Soofi S Pretraining Report v1.0") describes a ~30B (30B-A3B) hybrid Mamba-2/Transformer MoE.
Licenceread2026-11-19
The Soofi-S-Base Hugging Face page and Hub API show the repo gated "manual" and the licence tag "closed-beta" (license: other).
Model cardread2026-11-19
The Soofi-Project Hugging Face org lists 20 model repos, all gated "manual": the base, a Megatron-format base checkpoint (Soofi-S-Base-Megatron, created 2026-08-23), the Instruct, Isar and Rhine previews, and first-party GGUF, FP8 and EntQuant 2/3/4-bit builds of each preview.
Documentationread2026-11-19
The public soofi-project/Soofi-Pretraining GitHub repo (based on the Nvidia Nemotron 3 Nano architecture; last push 2026-07-15) says it holds "the complete codebase used for training Soofi S" (preprocessing, pretraining, midtraining).
Vendor announcementread2026-11-19
The project site soofi.info says "Soofi S befindet sich derzeit in der Beta- und Entwicklungsphase" and "Ein öffentlicher Open-Source-Release ist geplant", with no release date, no licence and no access-request process described.
Vendor announcementunverified2026-07-25
A consortium press release presents Soofi as a German public-private consortium project for sovereign industrial AI, trained on Deutsche Telekom's Industrial AI Cloud in Munich, coordinated by the KI Bundesverband and funded by the BMWE.
Third-party analysisunverified2026-07-25
CAIRNE's launch page describes the SOOFI consortium, its members (including CAIRNE Gold member L3S Research Centre) and its sovereign open-source foundation-model mission.
Third-party analysisunverified2026-07-25
Independent technology coverage reports Soofi-S as a leading fully-open model that meets the Open Source AI Definition, publishing weights, intermediate checkpoints, training and eval code and a detailed data inventory.
Third-party analysisunverified2026-07-25
Independent coverage reports Soofi-S is the strongest fully-open model on aggregate English and German benchmarks, ahead of OLMo 3 32B, Apertus 70B, EuroLLM 22B and Alia 40B, and winning code aggregates among open base models.