StandardThinking/StandardOne-8B-GGUF

Standard One 8B — GGUF

Updated weights (v2.2, 2026-10-04). These files are built from Standard One 8B v2.2. If you downloaded them before, download them again or pin a full commit revision. Earlier versions stay available under the tags v1, v1.1 and v2.

Decision API update (2026-10-08). All nine StandardOne-8B-*.gguf language model files now support stock llama.cpp's text /v1/systemone endpoint. Only five metadata fields were added; every tensor byte is unchanged. Download the updated files from main. The pre-update binaries remain at v2.2. mmproj is unchanged. See Decision API quick start.

Version: v2.2 weights + decision metadata (2026-10-08)

GGUF builds of Standard One 8B (Ministral 3 8B text + Pixtral vision tower, mistral3 architecture) for use with llama.cpp.

Files

Most quant levels below (marked "imatrix") were built with an importance matrix calibrated on 1,512 prompts sampled from our own training rows (see "Importance-matrix calibration" below), which recovers some of the accuracy quantization would otherwise lose. Q8_0 and BF16 don't need one.

File	Quant	Size	imatrix
StandardOne-8B-BF16.gguf	BF16 (no quantization)	17.0 GB	—
StandardOne-8B-Q8_0.gguf	Q8_0	9.0 GB	no
StandardOne-8B-Q5_K_M.gguf	Q5_K_M	6.1 GB	yes
StandardOne-8B-Q4_K_M.gguf	Q4_K_M	5.2 GB	yes
StandardOne-8B-IQ4_XS.gguf	IQ4_XS	4.7 GB	yes
StandardOne-8B-Q3_K_M.gguf	Q3_K_M	4.2 GB	yes
StandardOne-8B-IQ3_M.gguf	IQ3_M	4.0 GB	yes
StandardOne-8B-Q2_K.gguf	Q2_K	3.4 GB	yes
StandardOne-8B-IQ2_M.gguf	IQ2_M	3.1 GB	yes
mmproj-StandardOne-8B.gguf	F16 vision projector	857 MB	—

SHA256 checksums: SHA256SUMS. Source revisions, conversion tool version and full validation numbers: release-manifest.json.

Q4_K_M note: this is the imatrix-calibrated version, not a plain quantization. We generated both and chose whichever scored higher on the mean of 10 held-out and public-dataset decision suites (no JevBench items; imatrix 76.56 vs. plain 76.33; measured on an earlier version). v2.2 keeps the imatrix build.

Decision API quick start

Use any of the nine language model files in this repository; Q4_K_M is used below as an example. The update uses llama.cpp's existing openjev letter-logit readout with a Standard One native template. It is still Standard One weights. No engine fork, adapter server, or new weight conversion is needed. To use another precision, replace the Q4_K_M filename in the download, checksum verification, and server commands with the selected filename from the table. All nine files use the same endpoint.

1. Install and download

Use the tested official llama.cpp b11495 release for your platform. Extract the runtime, retain its companion libraries, and put llama-server on PATH (or use the full executable path). This release reports 0.6.0-dev, commit 37ac63456. Older builds may lack the decision endpoint; newer builds were not tested here.

In a Python 3.10+ virtual environment, install the Hub CLI:

python -m pip install --upgrade huggingface_hub
hf download StandardThinking/StandardOne-8B-GGUF \
  StandardOne-8B-Q4_K_M.gguf SHA256SUMS release-manifest.json \
  --revision main --local-dir models/standardone-8b-systemone

These shell examples use Bash/Zsh. The repository is public; login is unnecessary. For a reproducible snapshot, replace main with a full commit hash from Files and versions. The v2.2 tag contains the original binaries without decision metadata. Use main or the metadata-update commit for this endpoint.

Verify just the downloaded model (the other entries in SHA256SUMS are optional):

python - <<'PYVERIFY'
import hashlib
from pathlib import Path
root = Path("models/standardone-8b-systemone")
name = "StandardOne-8B-Q4_K_M.gguf"
expected = {}
for line in (root / "SHA256SUMS").read_text().splitlines():
    if line.strip():
        digest, filename = line.split(maxsplit=1)
        expected[filename.lstrip("*")] = digest
h = hashlib.sha256()
with (root / name).open("rb") as stream:
    for chunk in iter(lambda: stream.read(8 * 1024 * 1024), b""):
        h.update(chunk)
if h.hexdigest() != expected.get(name):
    raise SystemExit("SHA256 mismatch")
print("SHA256 OK")
PYVERIFY

2. Run the local decision endpoint

llama-server \
  -m models/standardone-8b-systemone/StandardOne-8B-Q4_K_M.gguf \
  -ngl 0 -t 4 -tb 4 -c 4096 -np 1 --jinja \
  --host 127.0.0.1 --port 8080 --alias standardone-8b

This is the tested CPU configuration. The model file is about 5.2 GB; RAM use also includes the KV cache and working buffers. Keep the server running. In another terminal, wait until curl --fail http://127.0.0.1:8080/health returns {"status":"ok"}. Stop the server with Ctrl+C when finished. GPU offload and image input were not validated for this decision API.

3. Send Jev-style typed questions

curl --fail http://127.0.0.1:8080/v1/systemone \
  -H 'Content-Type: application/json' \
  --data-binary @- <<'JSON'
{
  "state": "A parcel has a large tear and its contents are visible.",
  "questions": {
    "route": {
      "type": "choice",
      "instructions": "Should this parcel pass inspection?",
      "criteria": {
        "pass": "The packaging is intact.",
        "review": "The packaging is visibly damaged.",
        "unclear": "There is not enough information."
      }
    },
    "damaged": {
      "type": "noul",
      "instructions": "Is the packaging visibly damaged?",
      "criteria": {
        "true": "A large tear is visible.",
        "false": "The packaging is intact."
      }
    },
    "damage": {
      "type": "score",
      "instructions": "How damaged is the packaging?",
      "criteria": ["intact", "minor cosmetic damage", "large tear with visible contents"]
    }
  }
}
JSON

Read answers.route.choice and answers.route.probabilities, answers.damaged.noul (probability of yes), and answers.damage.score (expected zero-based level). score can be fractional. The server constructs JSON from option logits; the model does not generate a letter or JSON text. usage.output_tokens is 0. A downloadable request is in systemone-request.json.

The request demonstrates the response fields. Values differ across model sizes and quantizations; inspect the live response rather than treating one fixture as an accuracy guarantee.

What was verified, and what differs from the merged model's served endpoint

Each of the nine files passed six real-model fixtures: choice, noul, score, object state, reversed lexical choice-key order, and all three types in one request. Direct option-logprob readout agreed within 5.27e-09 for single-question fixtures and 4.44e-09 for the multi-question fixture. Multi-question reference calls follow the same question order and reuse prefix state after the first question. Independently prefilled references can differ due to CPU batching and cache behavior. Per-file results and tolerances are in systemone-validation.json. These checks verify endpoint calculations, not task accuracy.

• This API uses the native prompt, with temperatures choice=0.85, noul=0.85, score=0.70. They are the published native preset carried forward from v2 fitting, not temperatures fitted on these quantized weights.
• The current merged v2.2 default endpoint uses served wording, temperature 1.65, and extended uppercase labels. This implementation does not reproduce that preset.
• Supply at most 26 options. The embedded template rejects more; do not infer Standard One support for the underlying openjev path's larger label count.
• Text state, string instructions, string choice descriptions, and string score levels are the tested path. Object state is supported by this template, but preserves submitted key order; the existing adapter's native state renderer sorts keys. Keep state and option ordering fixed for comparisons.
• llama.cpp's confidence formulas differ from the existing adapter's normalized entropy. Matching probability fields does not imply identical confidence semantics.
• All nine 8B language files were tested on CPU with text requests. GPU execution, images, structured question/description values, and complete request-option/error parity are unverified. Do not send images to this text-only template.

The template is published as systemone-native-template.jinja. The release manifest records every added metadata field and the identical tensor-data hash. Current checksums reflect the added metadata. Original binaries and checksums remain in the v2.2 snapshot; the manifest also records every source-file SHA256. For the default BF16 endpoint, use the merged model's serving instructions.

Importance-matrix calibration

Q5_K_M down to IQ2_M were quantized with llama-imatrix calibrated on 1,512 prompts (spread evenly across 72 training-data cohorts; training data only — no benchmark/held-out file was used), context 2048. Q8_0 and BF16 don't use an imatrix (high enough precision that it doesn't move the needle).

Historical weight accuracy validation

The following measurements predate the metadata update. Tensor data is unchanged; these are not a new benchmark of the native decision endpoint.

CPU check of v2.2 with llama.cpp llama-server (no GPU offload): accuracy of the most probable option label at the answer position, native prompt wording, GGUF's own chat template, one option order, measured 2026-10-04. BF16-GGUF, Q8_0, Q4_K_M and IQ2_M were scored on all suites below; the other quants on the two JevBench tiers. Score changes from v2.1 to v2.2 are listed in the StandardOne-8B card.

Quant	Easy (48)	Original (72)	Judge proxy (60)	Realistic (60)	Hard proxy (80)	Same answer as BF16-GGUF
BF16-GGUF	100.00	97.22	90.00	68.33	56.25	—
Q8_0	100.00	98.61	90.00	68.33	56.25	99.06 % (n=320)
Q5_K_M	100.00	97.22	—	—	—	100.00 % (n=120)
Q4_K_M (shipped, imatrix)	100.00	95.83	91.67	66.67	50.00	94.69 % (n=320)
IQ4_XS	100.00	98.61	—	—	—	99.17 % (n=120)
Q3_K_M	100.00	93.06	—	—	—	97.50 % (n=120)
IQ3_M	100.00	88.89	—	—	—	93.33 % (n=120)
Q2_K	100.00	94.44	—	—	—	95.00 % (n=120)
IQ2_M	100.00	87.50	86.67	65.00	45.00	86.88 % (n=320)

Note on the mmproj conversion

llama.cpp's stock --mmproj converter (as of the commit used here) drops the [IMG_BREAK] token embedding for HF-format Mistral3ForConditionalGeneration checkpoints (a filter meant to strip text-model tensors also strips the one row of the text embedding matrix the vision projector needs), so the mmproj file it produces fails to load in llama-server/llama-cli ("unable to find tensor v.token_embd.img_break"). The mmproj file in this folder was built with a small local patch that lets that one tensor through; see release-manifest.json -> known_issues_fixed for details. It loads and runs correctly with --mmproj.

Update history

2026-10-08 — llama.cpp decision API metadata

• Added metadata for stock llama.cpp's /v1/systemone endpoint to all nine language GGUF files: BF16, Q8_0, Q5_K_M, Q4_K_M, IQ4_XS, Q3_K_M, IQ3_M, Q2_K, and IQ2_M.
• Added the decision type, a named native decision template, and per-type temperatures. The v2.2 weights and every tensor byte are unchanged; the vision projector (mmproj) is unchanged.
• Verified choice, noul, and score with official llama.cpp b11495 on every precision: 54 API fixture requests for this repository, including object state, option order, and multi-question requests. See systemone-validation.json.
• Added decision endpoint instructions and request examples, and updated SHA256SUMS and release-manifest.json. This implementation uses native wording with at most 26 options and text input; it does not reproduce the merged model's default served preset.
• Pre-update GGUF binaries remain available at v2.2. Metadata update commit.

License

Apache License 2.0 — see LICENSE and NOTICE. Same terms as the source StandardOne-8B release; this GGUF conversion adds no additional restrictions.