juspay

xor

README

License: apache-2.0

Base model

Table with columns: Field, Value
Field	Value
Base checkpoint	`Qwen/Qwen3.6-35B-A3B`
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 also include an images array containing up to eight image URLs or data URLs. Typed questions can classify information from the supplied images and text

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

The following commands download the validated release, verify and extract the serving bundle, and start Xor on two GPUs:

bash
export XOR_REVISION=xor-v1
export INSTALL_ROOT="$PWD/xor-eval"
export MODEL_DIR="$INSTALL_ROOT/model"

hf download juspay/xor   --revision "$XOR_REVISION"   --local-dir "$MODEL_DIR"

(
  cd "$MODEL_DIR/serving"
  sha256sum -c xor-serving.tar.gz.sha256
)

tar -tzf "$MODEL_DIR/serving/xor-serving.tar.gz"

mkdir -p "$INSTALL_ROOT/runtime"
tar -xzf "$MODEL_DIR/serving/xor-serving.tar.gz"   -C "$INSTALL_ROOT/runtime"   --strip-components=1

cd "$INSTALL_ROOT/runtime"
export CUDA_VISIBLE_DEVICES=0,1
export TP_SIZE=2
export API_PORT=49001
./run.sh

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

When the smoke test succeeds, the API is available at http://127.0.0.1:49001/v1/systemone

Test it with the included request:

bash
curl -sS -X POST http://127.0.0.1:49001/v1/systemone   -H 'Content-Type: application/json'   --data @examples/request.json

Test image input with the included data-URL request:

bash
curl -sS -X POST http://127.0.0.1:49001/v1/systemone \
  -H 'Content-Type: application/json' \
  --data @examples/image-request.json

Public JEVBench self-run

Xor was evaluated locally on the public tiers of JEVBench using harness commit fd51755eb0c0b546ca206d764faf3302feca913e, the existing typesafe adapter, and one request at a time

The run used 2 x NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs with 96 GB each, tensor parallelism 2, and the pinned serving image listed below

Table with columns: Tier, Attempted, Valid, Correct, Accuracy, Macro accuracy, Brier mean, ECE, p50, p95
Tier	Attempted	Valid	Correct	Accuracy	Macro accuracy	Brier mean	ECE	p50	p95
Easy	48	48	48	1.0000	1.0000	0.0018	0.0257	0.0766 s	0.0888 s

Operational success, coverage, schema validity, and strict schema validity were 1.0000 for all three public tiers

These are self-run public-tier results, not an official JEVBench rank. An independent full-suite evaluation, including held-out items, has been requested in JEVBench issue #20

Validated runtime

Table with columns: Setting, Value
Setting	Value
SGLang image	`lmsysorg/sglang@sha256:6bcaa47db52f78ce0d67863b8b2431221b79bc23204a80cad757fa819d00e921`
Tensor parallelism	2
Validated GPUs	2 x NVIDIA RTX PRO 6000 Blackwell Server Edition, 96 GB each
Maximum prefill tokens	250,000
Static memory fraction	0.85

For directly comparable latency measurements, use the same GPU model, GPU count, tensor-parallel configuration, and pinned serving image. Latency from other hardware should be identified as hardware-specific

Operational notes

The model files occupy approximately 66 GB. The validated SGLang configuration reserves substantial GPU memory for its KV cache. Smaller prefill budgets may work on other two-GPU configurations but are not covered by the validated setup

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

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Model Details

Model Provider

juspay

Model Tree

Base

Qwen/Qwen3.6-35B-A3B

Fine-tuned

this model

Input Modalities

TextImageVideo

Output Modalities

Text

Supported Functionality