Total runs:
24.5K
Run Growth:
491
Growth Rate:
2.01%
Sponsored by
YouArt
arcee-ai / Trinity-Nano-Preview-NVFP4
Model's Last Updated: March 26 2026
text-generation
Introduction of Trinity-Nano-Preview-NVFP4
Model Details of Trinity-Nano-Preview-NVFP4
Trinity Nano Preview NVFP4
Trinity Nano Preview is a preview of Arcee AI's 6B MoE model with 1B active parameters. It is the small-sized model in our new Trinity family, a series of open-weight models for enterprise and tinkerers alike.
This is a chat tuned model, with a delightful personality and charm we think users will love. We note that this model is pushing the limits of sparsity in small language models with only 800M non-embedding parameters active per token, and as such may be unstable in certain use cases, especially in this preview.
This is an experimental release, it's fun to talk to but will not be hosted anywhere, so download it and try it out yourself!
Trinity Nano Preview is trained on 10T tokens gathered and curated through a key partnership with Datology , building upon the excellent dataset we used on AFM-4.5B with additional math and code.
Training was performed on a cluster of 512 H200 GPUs powered by Prime Intellect using HSDP parallelism.
More details, including key architecture decisions, can be found on our blog here
This repository contains the NVFP4 quantized weights of Trinity-Nano-Preview for deployment on NVIDIA Blackwell GPUs.
Model Details
- Model Architecture: AfmoeForCausalLM
- Parameters: 6B, 1B active
- Experts: 128 total, 8 active, 1 shared
- Context length: 128k
- Training Tokens: 10T
- License: Apache 2.0
Quantization Details
-
Scheme:
NVFP4 (
nvfp4_mlp_only— MLP/expert weights only, attention remains BF16) - Tool: NVIDIA ModelOpt
- Calibration: 512 samples, seq_length=2048, all-expert calibration enabled
- KV cache: Not quantized
Running with vLLM
Requires vLLM >= 0.15.1.
vllm serve arcee-ai/Trinity-Nano-Preview-NVFP4 \
--quantization modelopt \
--trust-remote-code \
--enforce-eager \
--gpu-memory-utilization 0.90 \
--max-model-len 8192 \
--host 0.0.0.0 \
--port 8000
Note (Blackwell GPUs): As of vLLM 0.18.x, the native Blackwell FP4 MoE kernels (CUTLASS/FlashInfer) may not support all MoE architectures. If you encounter GEMM initialization errors or empty output on B200/B300/GB300, use the Marlin MoE backend:
export VLLM_NVFP4_GEMM_BACKEND=marlin
vllm serve arcee-ai/Trinity-Nano-Preview-NVFP4 \
--trust-remote-code \
--moe-backend marlin \
--gpu-memory-utilization 0.90 \
--max-model-len 8192 \
--host 0.0.0.0 \
--port 8000
Marlin decompresses FP4 weights to BF16 for compute, providing the full memory compression benefit (~3.7× vs BF16) but not native FP4 compute speedup. On Hopper GPUs (H100/H200), Marlin is selected automatically and no extra flags are needed. We are working with the vLLM team on native FP4 kernel support.
License
Trinity-Nano-Preview-NVFP4 is released under the Apache-2.0 license.
Runs of arcee-ai Trinity-Nano-Preview-NVFP4 on huggingface.co
351
Total runs
0
24-hour runs
4
3-day runs
26
7-day runs
351
30-day runs
More Information About Trinity-Nano-Preview-NVFP4 huggingface.co Model
Url of Trinity-Nano-Preview-NVFP4
Trinity-Nano-Preview-NVFP4 huggingface.co Url
Provider of Trinity-Nano-Preview-NVFP4 huggingface.co
Other API from arcee-ai
Updated:May 29 2026
Total runs:
10.0K
Run Growth:
9.3K
Growth Rate:
92.47%
Total runs:
4.2K
Run Growth:
-6.3K
Growth Rate:
-147.69%
Total runs:
4.0K
Run Growth:
-11.5K
Growth Rate:
-285.62%
Total runs:
1.8K
Run Growth:
1.5K
Growth Rate:
82.87%
Total runs:
1.7K
Run Growth:
-860
Growth Rate:
-49.60%
Total runs:
1.5K
Run Growth:
1.5K
Growth Rate:
98.63%
Total runs:
1.5K
Run Growth:
427
Growth Rate:
30.90%
Total runs:
1.2K
Run Growth:
-106
Growth Rate:
-8.55%
Total runs:
1.1K
Run Growth:
254
Growth Rate:
26.62%
Total runs:
1.1K
Run Growth:
729
Growth Rate:
79.50%
Total runs:
1.1K
Run Growth:
-30
Growth Rate:
-3.21%
Total runs:
1.0K
Run Growth:
-1.8K
Growth Rate:
-174.58%
Total runs:
850
Run Growth:
0
Growth Rate:
0.00%
Total runs:
850
Run Growth:
-607
Growth Rate:
-65.69%
Total runs:
766
Run Growth:
-6.3K
Growth Rate:
-820.89%
Total runs:
643
Run Growth:
431
Growth Rate:
69.85%
Total runs:
626
Run Growth:
-777
Growth Rate:
-121.60%
Total runs:
614
Run Growth:
-4.4K
Growth Rate:
-715.15%
Total runs:
528
Run Growth:
-163
Growth Rate:
-30.87%
Total runs:
505
Run Growth:
201
Growth Rate:
39.80%
Total runs:
503
Run Growth:
382
Growth Rate:
75.94%
Total runs:
486
Run Growth:
-26
Growth Rate:
-8.02%
Total runs:
448
Run Growth:
107
Growth Rate:
23.88%
Total runs:
408
Run Growth:
0
Growth Rate:
0.00%
Total runs:
281
Run Growth:
0
Growth Rate:
0.00%
Total runs:
258
Run Growth:
23
Growth Rate:
8.91%
Total runs:
256
Run Growth:
199
Growth Rate:
92.56%
Total runs:
243
Run Growth:
-1.7K
Growth Rate:
-680.25%
Total runs:
242
Run Growth:
202
Growth Rate:
84.17%
Total runs:
228
Run Growth:
-833
Growth Rate:
-387.44%
Total runs:
183
Run Growth:
142
Growth Rate:
78.45%
Total runs:
177
Run Growth:
98
Growth Rate:
55.06%
Total runs:
176
Run Growth:
105
Growth Rate:
60.69%
Total runs:
170
Run Growth:
-59
Growth Rate:
-36.88%
Total runs:
158
Run Growth:
-1
Growth Rate:
-0.63%
Total runs:
148
Run Growth:
-99
Growth Rate:
-66.89%
Total runs:
141
Run Growth:
105
Growth Rate:
77.21%
Total runs:
129
Run Growth:
71
Growth Rate:
56.80%
Total runs:
120
Run Growth:
14
Growth Rate:
11.29%
Total runs:
115
Run Growth:
61
Growth Rate:
54.46%
Total runs:
104
Run Growth:
78
Growth Rate:
78.79%
Total runs:
98
Run Growth:
88
Growth Rate:
92.63%
Total runs:
97
Run Growth:
80
Growth Rate:
86.02%
Total runs:
97
Run Growth:
86
Growth Rate:
91.49%
Total runs:
94
Run Growth:
71
Growth Rate:
78.89%
Total runs:
93
Run Growth:
42
Growth Rate:
45.65%
Total runs:
92
Run Growth:
67
Growth Rate:
74.44%
Total runs:
92
Run Growth:
33
Growth Rate:
35.87%
Total runs:
91
Run Growth:
66
Growth Rate:
76.74%
Total runs:
91
Run Growth:
78
Growth Rate:
86.67%
Total runs:
91
Run Growth:
78
Growth Rate:
89.66%
Total runs:
90
Run Growth:
76
Growth Rate:
86.36%
Total runs:
90
Run Growth:
77
Growth Rate:
88.51%
Total runs:
90
Run Growth:
74
Growth Rate:
85.06%
Total runs:
90
Run Growth:
76
Growth Rate:
86.36%
Total runs:
89
Run Growth:
69
Growth Rate:
79.31%
Total runs:
88
Run Growth:
80
Growth Rate:
93.02%
Total runs:
88
Run Growth:
79
Growth Rate:
89.77%
Total runs:
88
Run Growth:
81
Growth Rate:
93.10%
Total runs:
87
Run Growth:
75
Growth Rate:
91.46%
Total runs:
87
Run Growth:
75
Growth Rate:
90.36%
Total runs:
86
Run Growth:
77
Growth Rate:
91.67%
Total runs:
86
Run Growth:
64
Growth Rate:
79.01%
Total runs:
86
Run Growth:
65
Growth Rate:
76.47%
Total runs:
86
Run Growth:
77
Growth Rate:
89.53%
Total runs:
86
Run Growth:
76
Growth Rate:
91.57%
Total runs:
85
Run Growth:
69
Growth Rate:
83.13%
Total runs:
85
Run Growth:
74
Growth Rate:
89.16%
Total runs:
85
Run Growth:
73
Growth Rate:
87.95%
Total runs:
85
Run Growth:
75
Growth Rate:
91.46%
Total runs:
84
Run Growth:
76
Growth Rate:
92.68%