ndc
NDC
25 results
VM Types
—
Engines
vllm
GPUs
NVIDIA GeForce RTX 3090, NVIDIA RTX PRO 6000 Blackwell Max-Q
Compute
GPU
everything above the entry cards — consumer Ampere–Ada through Blackwell / Hopper datacenter silicon, any count; ranked by capability composite
#
Model
Hardware – Engine
Ships
Capability
Tiers
Agentic
Refusals
tok/s
Date
1
Qwen/Qwen3.8-27B-FP8 verified
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.77n=89
0.96±0.020
0.96 usability
0.95 agentic work
0.98 · 0.98 · 0.83
1.00
0
65.9
Aug 30
2
Qwen/Qwen3.8-27B-FP8 verified
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.77n=89
0.96±0.023
0.95 usability
0.93 agentic work
0.97 · 0.96 · 0.83
0.96
0
64.3
Aug 30
3
Qwen/Qwen3.8-27B-FP8 verified
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.77n=89
0.96±0.023
0.95 usability
0.93 agentic work
0.97 · 0.93 · 0.89
0.90
0
68.4
Sep 10
4
Qwen/Qwen3.8-27B-FP8 verified
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.77n=89
0.95±0.027
0.94 usability
0.93 agentic work
0.97 · 0.92 · 0.83
0.92
0
67.5
Sep 11
5
openai/gpt-oss-120b verified
4×NVIDIA GeForce RTX 3090 – vllm
ndc
0.73n=53
0.92±0.037
0.91 usability
0.91 agentic work
0.95 · 0.91 · 0.75
0.89
0
132.5
Aug 30
6
google/gemma-4-31B-it-qat-w4a16-ct verifiedno-think
Googleqat-w4a16
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=42
0.92±0.026
0.93 usability
0.94 agentic work
0.94 · 0.91 · 0.93
0.96
0
47.3
Sep 10
7
openai/gpt-oss-120b verified
4×NVIDIA GeForce RTX 3090 – vllm
ndc
0.73n=53
0.92±0.045
0.90 usability
0.92 agentic work
0.97 · 0.84 · 0.70
0.78
0
79.2
Sep 10
7
openai/gpt-oss-120b verified
4×NVIDIA GeForce RTX 3090 – vllm
ndc
0.73n=53
0.92±0.054
0.88 usability
0.92 agentic work
0.96 · 0.82 · 0.72
0.67
0
134.5
Aug 30
7
google/gemma-4-31B-it-qat-w4a16-ct verifiedno-think
Googleqat-w4a16
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=42
0.92±0.027
0.92 usability
0.95 agentic work
0.93 · 0.92 · 0.86
0.91
0
46.7
Aug 30
10
google/gemma-4-31B-it-qat-w4a16-ct verifiedno-think
Googleqat-w4a16
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=42
0.92±0.026
0.93 usability
0.94 agentic work
0.94 · 0.91 · 0.91
0.95
0
46.4
Sep 9
11
openai/gpt-oss-120b verified
4×NVIDIA GeForce RTX 3090 – vllm
ndc
0.73n=53
0.91±0.040
0.89 usability
0.89 agentic work
0.95 · 0.83 · 0.85
0.80
0
77.5
Sep 9
12
Qwen/Qwen3.8-27B-FP8 verified
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.77n=89
0.89±0.103
0.82 usability
0.74 agentic work
—
0.60
0
69.3
Aug 26
13
poolside/Laguna-S-2.1-NVFP4 verifiedno-think
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=85
0.89±0.045
0.89 usability
0.92 agentic work
0.94 · 0.89 · 0.61
0.96
0
98.7
Aug 30
13
deepseek-ai/DeepSeek-V4-Flash-0731 verifiedno-think
2×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.94n=50
0.89±0.048
0.89 usability
0.94 agentic work
0.95 · 0.85 · 0.69
0.97
0
57.4
Sep 10
15
deepseek-ai/DeepSeek-V4-Flash-0731 verifiedno-think
2×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.94n=50
0.89±0.053
0.88 usability
0.91 agentic work
0.94 · 0.88 · 0.58
1.00
0
58.5
Aug 30
16
deepseek-ai/DeepSeek-V4-Flash-0731 verifiedno-think
2×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.94n=50
0.88±0.050
0.89 usability
0.92 agentic work
0.95 · 0.85 · 0.73
0.97
0
58.2
Sep 9
17
poolside/Laguna-S-2.1-NVFP4 verifiedno-think
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=85
0.88±0.053
0.87 usability
0.91 agentic work
0.93 · 0.86 · 0.63
0.88
0
98.4
Sep 9
18
poolside/Laguna-S-2.1-NVFP4 verifiedno-think
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=85
0.87±0.050
0.85 usability
0.87 agentic work
0.93 · 0.83 · 0.58
0.80
0
98.6
Sep 10
19
poolside/Laguna-S-2.1-NVFP4 verifiedno-think
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=85
0.86±0.048
0.85 usability
0.88 agentic work
0.91 · 0.84 · 0.61
0.89
0
96.3
Aug 30
20
deepseek-ai/DeepSeek-V4-Flash-0731 verifiedno-think
2×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.94n=50
0.86±0.066
0.84 usability
0.91 agentic work
0.89 · 0.87 · 0.54
1.00
0
57.3
Aug 30
21
google/gemma-4-31B-it-qat-w4a16-ct verifiedno-think
Googleqat-w4a16
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=42
0.86±0.106
0.85 usability
0.96 agentic work
0.81 · 0.92 · 0.84
0.98
0
48.2
Aug 30
22
openai/gpt-oss-120b verified
4×NVIDIA GeForce RTX 3090 – vllm
ndc
0.73n=53
0.84±0.111
0.76 usability
0.70 agentic work
—
0.56
0
130.4
Aug 26
23
poolside/Laguna-S-2.1-NVFP4 verifiedno-think
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=85
0.81±0.100
0.76 usability
0.71 agentic work
—
0.69
0
93.6
Aug 26
24
deepseek-ai/DeepSeek-V4-Flash-0731 verifiedno-think
2×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.94n=50
0.80±0.105
0.73 usability
0.70 agentic work
—
0.60
0
57.1
Aug 26
25
google/gemma-4-31B-it-qat-w4a16-ct verifiedno-think
Googleqat-w4a16
1×NVIDIA RTX PRO 6000 Blackwell Max-Q – vllm
ndc
0.82n=42
0.79±0.140
0.71 usability
0.72 agentic work
—
0.57
0
47.4
Aug 26
1
0.96±0.020
capability
ndc · vllm99% done · 2 timeouts65.9 tok/s
Categories:
Aug 30
2
0.96±0.023
capability
ndc · vllm99% done · 2 timeouts64.3 tok/s
Categories:
Aug 30
3
0.96±0.023
capability
ndc · vllm99% done · 2 timeouts68.4 tok/s
Categories:
Sep 10
4
0.95±0.027
capability
ndc · vllm98% done · 4 timeouts67.5 tok/s
Categories:
Sep 11
5
openai/gpt-oss-120bGPU verified
OpenAI0.92±0.037
capability
ndc · vllm100% done132.5 tok/s
Categories:
Aug 30
6
0.92±0.026
capability
ndc · vllm100% done47.3 tok/s
Categories:
Sep 10
7
openai/gpt-oss-120bGPU verified
OpenAI0.92±0.045
capability
ndc · vllm99% done · 1 timeout79.2 tok/s
Categories:
Sep 10
8
openai/gpt-oss-120bGPU verified
OpenAI0.92±0.054
capability
ndc · vllm100% done134.5 tok/s
Categories:
Aug 30
9
0.92±0.027
capability
ndc · vllm100% done46.7 tok/s
Categories:
Aug 30
10
0.92±0.026
capability
ndc · vllm100% done46.4 tok/s
Categories:
Sep 9
11
openai/gpt-oss-120bGPU verified
OpenAI0.91±0.040
capability
ndc · vllm100% done77.5 tok/s
Categories:
Sep 9
12
0.89±0.103
capability
ndc · vllm99% done · 1 timeout69.3 tok/s
Categories:
Aug 26
13
poolside/Laguna-S-2.1-NVFP4GPU verifiedno-think
poolside0.89±0.045
capability
ndc · vllm100% done98.7 tok/s
Categories:
Aug 30
14
deepseek-ai/DeepSeek-V4-Flash-0731GPU verifiedno-think
DeepSeek0.89±0.048
capability
ndc · vllm100% done57.4 tok/s
Categories:
Sep 10
15
deepseek-ai/DeepSeek-V4-Flash-0731GPU verifiedno-think
DeepSeek0.89±0.053
capability
ndc · vllm100% done58.5 tok/s
Categories:
Aug 30
16
deepseek-ai/DeepSeek-V4-Flash-0731GPU verifiedno-think
DeepSeek0.88±0.050
capability
ndc · vllm99% done · 1 timeout58.2 tok/s
Categories:
Sep 9
17
poolside/Laguna-S-2.1-NVFP4GPU verifiedno-think
poolside0.88±0.053
capability
ndc · vllm100% done98.4 tok/s
Categories:
Sep 9
18
poolside/Laguna-S-2.1-NVFP4GPU verifiedno-think
poolside0.87±0.050
capability
ndc · vllm100% done98.6 tok/s
Categories:
Sep 10
19
poolside/Laguna-S-2.1-NVFP4GPU verifiedno-think
poolside0.86±0.048
capability
ndc · vllm99% done · 1 timeout96.3 tok/s
Categories:
Aug 30
20
deepseek-ai/DeepSeek-V4-Flash-0731GPU verifiedno-think
DeepSeek0.86±0.066
capability
ndc · vllm100% done57.3 tok/s
Categories:
Aug 30
21
0.86±0.106
capability
ndc · vllm100% done48.2 tok/s
Categories:
Aug 30
22
openai/gpt-oss-120bGPU verified
OpenAI0.84±0.111
capability
ndc · vllm100% done130.4 tok/s
Categories:
Aug 26
23
poolside/Laguna-S-2.1-NVFP4GPU verifiedno-think
poolside0.81±0.100
capability
ndc · vllm100% done93.6 tok/s
Categories:
Aug 26
24
deepseek-ai/DeepSeek-V4-Flash-0731GPU verifiedno-think
DeepSeek0.80±0.105
capability
ndc · vllm100% done57.1 tok/s
Categories:
Aug 26
25
0.79±0.140
capability
ndc · vllm100% done47.4 tok/s
Categories:
Aug 26