Xor 1.1

Xor 1.1 (xor-1.1) is a post-trained version of Qwen/Qwen3.6-35B-A3B for typed decision tasks. It is served through a TypeSafe-compatible /v1/systemone API.

Changes from Xor 1.0

• New post-trained weights (LoRA rank 16, merged into BF16 base weights)
• choice and score questions accept 2 to 255 candidates (previously 26)
• Video input in addition to images; remote media URLs are rejected, only data URLs are accepted
• Per-question-type calibration temperatures: noul 1.4, choice 1.1, score 1.0
• Validated on one GPU (TP1) with a pinned, patched SGLang runtime

Revisions

Revision	Release
main	Latest release (currently Xor 1.1)
v1.1	Xor 1.1, immutable
xor-v1	Xor 1.0, immutable

Pin a revision for reproducible results:

hf download juspay/xor --revision v1.1 --local-dir xor-1.1

Base model

Field	Value
Base checkpoint	Qwen/Qwen3.6-35B-A3B
Base revision	995ad96eacd98c81ed38be0c5b274b04031597b0
Architecture	Mixture-of-experts causal language model
Total parameters	35 billion
Activated parameters	Approximately 3 billion per token
Released precision	BF16
Base license	Apache License 2.0
Packaging	Fully merged weights; no adapter loading or merging is required

Interface

The server accepts a state and a map of typed questions:

• noul: binary probability
• choice: categorical decision and full probability distribution
• score: expected ordinal score and full probability distribution

Requests may include an images array with up to eight image data URLs, or one video data URL. Typed questions can classify information from the supplied media and text. Send one kind of media per request: the API accepts images and a video together, but the model does not reliably tell them apart.

The complete request body must not exceed 8 MB, which leaves roughly 6 MB of media after base64 encoding. Larger requests are rejected with HTTP 422.

The serving layer performs deterministic single-token candidate readout, forward and reverse option-order evaluation, probability calibration, and schema conversion. The serving layer is part of the released inference configuration and must be used for reproducible results.

Quick start

Download the release, verify and extract the serving bundle, and start Xor on one GPU:

hf download juspay/xor --revision v1.1 --local-dir xor-1.1

(cd xor-1.1/serving && sha256sum -c xor-1.1-serving.tar.gz.sha256)

mkdir -p xor-1.1-runtime
tar -xzf xor-1.1/serving/xor-1.1-serving.tar.gz -C xor-1.1-runtime --strip-components=1

cd xor-1.1-runtime
cp .env.example .env
sed -i "s|^MODEL_DIR=.*|MODEL_DIR=$(cd ../xor-1.1 && pwd)|" .env
./run.sh

run.sh verifies every model file against checksums.sha256 before starting. When the smoke test succeeds, the API is available at http://127.0.0.1:30002/v1/systemone.

curl -sS -X POST http://127.0.0.1:30002/v1/systemone \
  -H 'Content-Type: application/json' \
  --data @examples/request.json

curl -sS -X POST http://127.0.0.1:30002/v1/systemone \
  -H 'Content-Type: application/json' \
  --data @examples/image-request.json

The setup requires Linux x86-64, the Hugging Face CLI, Docker Engine with Docker Compose v2, the NVIDIA Container Toolkit, and approximately 120 GB of free disk space.

Public JEVBench self-run

Xor 1.1 was evaluated locally on the public tiers of JEVBench using harness commit 1bcc55eb6c8cffde2306b3db03ede39b61c6152a, the existing typesafe adapter, and one request at a time.

The run used the released serving bundle unmodified on 1 x NVIDIA H200 (143 GB), tensor parallelism 1, data parallelism 1, with request caching disabled.

Tier	Attempted	Valid	Correct	Accuracy	p50	p95
Easy	48	48	48	1.0000	0.0739 s	0.0772 s
Original	72	72	70	0.9722	0.0742 s	0.0780 s
Hard public	111	111	89	0.8018	0.0781 s	0.1378 s
All public	231	231	207	0.8961	0.0753 s	0.1207 s

Across all 231 public decisions: macro accuracy 0.9056, Brier mean 0.1704, ECE 0.0559. Operational success, coverage, schema validity, and strict schema validity were 1.0000.

These are self-run public-tier results, not an official JEVBench rank. Latency is hardware-specific and was measured locally without network overhead.

Validated runtime

Setting	Value
SGLang image	prakhar1611/xor-sglang@sha256:94c48d2a6cc98dc456cf93f723707ea7dd81dddfe1061e823b348d68bbe8158f
Upstream SGLang base	lmsysorg/sglang@sha256:6bcaa47db52f78ce0d67863b8b2431221b79bc23204a80cad757fa819d00e921
Tensor parallelism	1
Data parallelism	1
Validated GPU	1 x NVIDIA H200, 143 GB
Maximum prefill tokens	250,000
Static memory fraction	0.85

The runtime image adds a small fix to SGLang's candidate-logprob result handling that prevents a scheduler crash when scoring and plain generation requests share a batch. Dockerfile.sglang-logprob-fix in the serving bundle reproduces it from the upstream image.

Operational notes

The model files occupy approximately 66 GB. Each data-parallel worker loads a complete model replica. Alternative hardware and parallelism settings must be validated independently before publishing performance results.

Artifact hashes, the base revision, adapter and merge metadata, the runtime image, and known provenance gaps are recorded in RELEASE_PROVENANCE.json and checksums.sha256.

The service does not require an API key when bound to localhost. Remote deployments must add authentication, TLS, rate limits, and request-size limits at the ingress layer.