Alpha
Contact
Support · Soofi

Will it last?

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

This page is a projection of the one entry record, the Transparency factor that Support 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

Common problems & fixes

Preview pitfalls, not an exhaustive catalogue:

  • AutoModel fails to load or reports an unknown architecture: Soofi-S is a custom hybrid Mamba-2/MoE. Pass trust_remote_code=True and use a recent transformers build.
  • Tool calling is silently ignored: the runtime is not applying the native tool template. Run llama-server --jinja (llama.cpp) instead of a runtime whose template strips tools (e.g. Ollama).
  • The serving stack rejects the model: custom-architecture support is still maturing. Confirm your vLLM or llama.cpp version supports the hybrid architecture, or fall back to a first-party GGUF build.
  • Out-of-memory when loading weights: decode uses ~3B active params, but all ~30B must be resident. Use a first-party GGUF or 3-bit quant or multi-GPU; the near-constant cache means long context is not the cause.
  • Non-reproducible results or silent updates: a deployment floating on main changes under you. Pin an exact Hugging Face revision and verify checksums.

Versions, changelog & cadence

Soofi-S is on an early preview/beta cadence, not a settled semantic-version stream. Alongside Soofi-S-Base there are Instruct-Preview, Isar-Preview (reasoning) and Rhine-Preview variants, first-party GGUF and 3-bit quant repos, and named iteration checkpoints (e.g. iter_1056000, the final constant-annealing checkpoint used for evaluation). Versions are tracked as Hugging Face revisions on the Soofi-Project org, so the revision hash is the changelog anchor: pin it and diff against a newer revision when upgrading. The pretraining technical report documents architecture, data and evaluation for the release.

Community & support channels

Usage questions go to the model and dataset discussion tabs on the Soofi-Project Hugging Face org (ev-hf-org). Training-code and reproduction problems go to GitHub issues on soofi-project/Soofi-Pretraining, soofi-trainer and soofi-model-hosting (ev-github). Announcements come from the soofi.info project site and the CAIRNE launch page (ev-cairne), plus the KI-Bundesverband-coordinated consortium behind the project. There is no paid support tier; support is community and consortium based.

Tracked known issues

These come from model-card caveats, not a formal issue tracker, so coverage is partial.

  • Preview status: capability, safety and tooling are still stabilising.
  • Safety and privacy features are documented as incomplete, with no companion guard model.
  • Tool-call templates are inconsistent across runtimes (native under --jinja; omitted by some).
  • Runtime support is immature: not every serving stack and version handles the hybrid Mamba-2/MoE yet.

How this scores

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

2

TransparencyDo you know what it is: weights, training, behaviour, and legible terms?

Moderate

The report, the per-source data accounting and the training code are public and inspectable, which is more than most weight-only releases offer. The weights are not publicly inspectable (manual gate, closed beta) and the training code has no licence file, so transparency stays at moderate until the weights and licence are released.

How this scores (AOI sub-dimensions)
Provenance4/5how well we can trace and verify what went into the modelDistributed from the verified Soofi-Project org on Hugging Face in bf16 safetensors (plus first-party GGUF and 3-bit quants) with checksums.
Governance4/5how accountable and well-documented the publisher isAccountable, well-identified publisher: a named government-funded consortium coordinated by the KI Bundesverband with reputable research members, a detailed pretraining technical report and open GitHub repositories.
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.