Instructions to use trillionlabs/TxGravity-30B-A6B-Open with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use trillionlabs/TxGravity-30B-A6B-Open with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="trillionlabs/TxGravity-30B-A6B-Open")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("trillionlabs/TxGravity-30B-A6B-Open")
model = AutoModelForCausalLM.from_pretrained("trillionlabs/TxGravity-30B-A6B-Open", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use trillionlabs/TxGravity-30B-A6B-Open with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "trillionlabs/TxGravity-30B-A6B-Open"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "trillionlabs/TxGravity-30B-A6B-Open",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/trillionlabs/TxGravity-30B-A6B-Open

SGLang

How to use trillionlabs/TxGravity-30B-A6B-Open with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "trillionlabs/TxGravity-30B-A6B-Open" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "trillionlabs/TxGravity-30B-A6B-Open",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "trillionlabs/TxGravity-30B-A6B-Open" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "trillionlabs/TxGravity-30B-A6B-Open",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use trillionlabs/TxGravity-30B-A6B-Open with Docker Model Runner:
```
docker model run hf.co/trillionlabs/TxGravity-30B-A6B-Open
```

TxGravity-30B-A6B-Open

Model Summary

TxGravity-30B-A6B-Open is a therapeutics-focused, task-specialized model built on the Gravity-30B-A5B base. It predicts a broad range of therapeutic properties — small-molecule ADMET, toxicity, drug–target interaction, protein–protein and peptide–MHC interaction, and more — by converting the Therapeutic Data Commons (TDC) benchmark tasks into an instruction format for LLMs, in the same spirit as Google's TxGemma.

TxGravity is co-developed by SK Biopharmaceuticals (SKBP) and Trillion Labs.

It is trained on TDC tasks reformatted as single-turn instructions (instruction → answer), covering 57 tasks. Answers are formatted as (A)/(B) for binary classification or a normalized 000–1000 bin for regression.

This release is the open-license version, with any datasets restricted to non-commercial use excluded from training.

Property	Value
Total Parameters	29.56B
Active Parameters	~6B
Architecture	GravityMoE (DeepSeek-V3-compatible)
Layers	52
Routed / Shared Experts	64 (top-8) / 1
Context Length	8,192 tokens
Precision	bf16
Base model	Gravity-30B-A5B

⚠️ This is a task-specialized property predictor, not a general instruction-tuned or safety-aligned assistant. Its outputs are intended for the TDC-style therapeutic prediction prompts it was trained on. Predictions may be inaccurate, biased, or incomplete and must be independently verified before any experimental, clinical, or decision-making use.

Performance

Evaluated on the TDC therapeutic benchmark against a reproduced TxGemma-27B baseline. Over 60 comparable tasks: 16 wins / 26 ties / 18 losses (6 tasks excluded for lack of a paper-reported comparison). A win/loss requires the margin to exceed the tie threshold; otherwise the task is scored a tie.

Wins (16 tasks) — TxGravity beats TxGemma-27B on: bindingdb_ki, bindingdb_ic50, protein_sabdab, ppbr_az, caco2_wang, lipophilicity_astrazeneca, half_life_obach, clearance_hepatocyte_az, ld50_zhu, bioavailability_ma, buchwald_hartwig, drugcomb_css, drugcomb_loewe, drugcomb_zip, drugcomb_hsa, drugcomb_bliss — spanning binding affinity, ADMET, toxicity, and drug-combination synergy.

Highlights: bindingdb_ki PCC 0.750 vs -0.112, bindingdb_ic50 Spearman 0.766 vs 0.643, protein_sabdab MAE 0.969 vs 2.332, ld50_zhu MAE 0.651 vs 0.776.

Full results

Margin is TxGravity - TxGemma, in percentage points (%p) for score metrics or relative percent (%rel) for error metrics (MAE/MSE, lower is better). Up-arrow = higher is better, down-arrow = lower is better.

Task	Metric	N	TxGravity	TxGemma-27B	Margin	Verdict
bindingdb_ic50	Spearman ↑	35725	0.7660	0.6430	+12.3%p	WIN
bindingdb_ki	PCC ↑	11857	0.7500	-0.1120	+86.2%p	WIN
bioavailability_ma	AUROC ↑	128	0.7490	0.6970	+5.2%p	WIN
buchwald_hartwig	PCC ↑	791	0.8920	0.8550	+3.7%p	WIN
caco2_wang	MAE ↓	182	0.4160	0.4660	+10.7%rel	WIN
clearance_hepatocyte_az	Spearman ↑	243	0.3240	0.2910	+3.3%p	WIN
drugcomb_bliss	MAE ↓	59708	3.7370	3.9280	+4.9%rel	WIN
drugcomb_css	MAE ↓	59708	7.8380	9.4180	+16.8%rel	WIN
drugcomb_hsa	MAE ↓	59708	3.5840	3.7880	+5.4%rel	WIN
drugcomb_loewe	MAE ↓	59708	6.4130	7.6720	+16.4%rel	WIN
drugcomb_zip	MAE ↓	59708	3.0610	3.4100	+10.2%rel	WIN
half_life_obach	Spearman ↑	135	0.3600	0.2890	+7.1%p	WIN
ld50_zhu	MAE ↓	1478	0.6510	0.7760	+16.1%rel	WIN
lipophilicity_astrazeneca	MAE ↓	840	0.5700	0.5950	+4.2%rel	WIN
ppbr_az	MAE ↓	559	8.2900	9.4180	+12.0%rel	WIN
protein_sabdab	MAE ↓	99	0.9690	2.3320	+58.4%rel	WIN
ames	AUROC ↑	1457	0.8410	0.8300	+1.1%p	TIE
bbb_martins	AUROC ↑	406	0.9190	0.8970	+2.2%p	TIE
bindingdb_patent	PCC ↑	49028	0.5410	0.5140	+2.7%p	TIE
carcinogens_lagunin	Accuracy ↑	56	0.8750	0.8930	-1.8%p	TIE
clearance_microsome_az	Spearman ↑	221	0.5000	0.4870	+1.3%p	TIE
clintox	AUROC ↑	297	0.7610	0.7560	+0.5%p	TIE
cyp1a2_veith	AUPRC ↑	2517	0.9310	0.9350	-0.4%p	TIE
cyp2c19_veith	AUROC ↑	2534	0.8940	0.8910	+0.3%p	TIE
cyp2c9_veith	AUPRC ↑	2419	0.7920	0.7820	+1.0%p	TIE
cyp2d6_veith	AUPRC ↑	2626	0.6840	0.6620	+2.2%p	TIE
cyp3a4_veith	AUPRC ↑	2467	0.8610	0.8410	+2.0%p	TIE
dili	AUROC ↑	96	0.8930	0.8890	+0.4%p	TIE
herg	AUROC ↑	132	0.8930	0.9000	-0.7%p	TIE
herg_central	AUROC ↑	61379	0.8760	0.8830	-0.7%p	TIE
herg_karim	Accuracy ↑	2690	0.7920	0.8000	-0.8%p	TIE
hia_hou	AUROC ↑	117	0.9860	0.9860	+0.0%p	TIE
hiv	AUROC ↑	8227	0.7640	0.7930	-2.9%p	TIE
huri	AUPRC ↑	4204	0.8080	0.7920	+1.6%p	TIE
mhc2_iedb_jensen	AUROC ↑	26856	0.8680	0.8500	+1.8%p	TIE
pgp_broccatelli	AUROC ↑	245	0.9380	0.9320	+0.6%p	TIE
sarscov2_3clpro_diamond	AUROC ↑	176	0.7470	0.7250	+2.2%p	TIE
skin_reaction	AUROC ↑	82	0.6510	0.6370	+1.4%p	TIE
tox21	AUROC ↑	15600	0.8230	0.8430	-2.0%p	TIE
toxcast	AUROC ↑	307282	0.8970	0.9050	-0.8%p	TIE
vdss_lombardo	Spearman ↑	226	0.5820	0.5660	+1.6%p	TIE
weber	AUROC ↑	9417	0.7180	0.7390	-2.1%p	TIE
bindingdb_kd	PCC ↑	1630	0.5220	0.5960	-7.4%p	LOSS
butkiewicz	AUROC ↑	401997	0.7540	0.8620	-10.8%p	LOSS
cyp2c9_substrate_carbonmangels	AUPRC ↑	135	0.3460	0.4730	-12.7%p	LOSS
cyp2d6_substrate_carbonmangels	AUPRC ↑	135	0.7080	0.7380	-3.0%p	LOSS
cyp3a4_substrate_carbonmangels	AUROC ↑	135	0.6410	0.6950	-5.4%p	LOSS
davis	MSE ↓	1064	0.7600	0.6430	-18.2%rel	LOSS
disgenet	MAE	10495	—	—	—	excluded
gdsc1	PCC	35462	—	—	—	excluded
gdsc2	PCC	18541	—	—	—	excluded
kiba	MSE ↓	4537	0.6370	0.4600	-38.5%rel	LOSS
leenay	Spearman ↑	1520	0.1510	0.2260	-7.5%p	LOSS
mhc1_iedb_imgt_nielsen	AUROC ↑	37197	0.9200	0.9680	-4.8%p	LOSS
mirtarbase	Accuracy ↑	160033	0.5550	0.8040	-24.9%p	LOSS
oncopolypharmacology	PCC ↑	4647	0.5050	0.5690	-6.4%p	LOSS
pampa_ncats	AUROC ↑	408	0.6630	0.7080	-4.5%p	LOSS
phase1	AUROC	561	—	—	—	excluded
phase2	AUROC	1279	—	—	—	excluded
phase3	AUROC	1200	—	—	—	excluded
sabdab_chen	AUPRC ↑	482	0.6580	0.7080	-5.0%p	LOSS
sarscov2_vitro_touret	AUROC ↑	298	0.4580	0.5300	-7.2%p	LOSS
solubility_aqsoldb	MAE ↓	1997	0.8550	0.8050	-6.2%rel	LOSS
tap	MAE ↓	240	5.4400	4.7850	-13.7%rel	LOSS
uspto	Accuracy ↑	221648	0.0000	0.0880	-8.8%p	LOSS
uspto_yields	PCC ↑	170728	-0.0571	0.1860	-24.3%p	LOSS

Also included in this repo as summary.csv.

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("trillionlabs/TxGravity-30B-A6B-Open")
model = AutoModelForCausalLM.from_pretrained(
    "trillionlabs/TxGravity-30B-A6B-Open", dtype="bfloat16", device_map="auto")

messages = [{"role": "user", "content": "<TDC-formatted instruction here>"}]
ids = tok.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
out = model.generate(ids, max_new_tokens=16)
print(tok.decode(out[0][ids.shape[1]:], skip_special_tokens=True))

The model expects TDC-style prompts and answers in TDC format: (A)/(B) for classification, or a 000–1000 normalized bin for regression.

License

Apache 2.0

Citation

@misc{txgravity2026,
  title        = {TxGravity-30B-A6B-Open},
  author       = {SK Biopharmaceuticals and Trillion Labs},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/trillionlabs/TxGravity-30B-A6B-Open}}
}

This model builds on the Therapeutic Data Commons (TDC) benchmark:

@article{huang2021therapeutics,
  title   = {Therapeutics Data Commons: Machine Learning Datasets and Tasks for Drug Discovery and Development},
  author  = {Huang, Kexin and Fu, Tianfan and Gao, Wenhao and Zhao, Yue and Roohani, Yusuf and Leskovec, Jure and Coley, Connor W and Xiao, Cao and Sun, Jimeng and Zitnik, Marinka},
  journal = {Proceedings of Neural Information Processing Systems, NeurIPS Datasets and Benchmarks},
  year    = {2021}
}

Downloads last month: -

Safetensors

Model size

30B params

Tensor type

BF16