Created bymario-alka.dePowered bygodcore.denoob2claw.detricoma.de
DEEN
Sign in Sign up
Beta
💰
LLM Economics

LLM Economics

Compare hardware, models and inference setups by operating cost, token output and total cost of ownership.

💸 Operating cost🖥️ TCO⚡ Tokens / euro🔋 Tokens / kWh
📊 Overview🎮 GPUs🧠 Models⚖️ Compare📚 Methodology
Reset

Settings are stored in your browser and shared via the URL. Power draw is estimated as a TDP sum since no per-run measurement exists. Methodology →

Model economics

52 Models

Most economical setup per model. No quality judgement – only measured cost per token. Number after the model = runs · HW configs.

#ModelMakerParamsGen tok/s iEUR / 1M tok iTok/€Tok/kWhcheapest HW
1 MiniMax-M2.7-AWQ-4biti MiniMax 230B 86.6 EUR 0.067 14.93M 4.48M NVIDIA GB10 (DGX Spark)
2 Meta-Llama-3.1-8B-Instructi Meta 8B 66.7 EUR 0.23 4.26M 1.28M AMD Radeon 8060S Graphics
3 Nemotron-Cascade-2-30B-A3Bi NVIDIA 30B 145.7 EUR 0.17 5.85M 1.75M AMD Radeon 8060S Graphics
4 Qwen3-VL-30B-A3B-Instructi Qwen (Alibaba) 30B 326.5 EUR 0.067 14.87M 4.46M NVIDIA RTX PRO 6000 Blackwell Workstation Edition
5 Nemotron-3-Nano-Omni-30B-A3B-Reasoningi NVIDIA 30B 144.1 EUR 0.17 5.74M 1.72M AMD Radeon 8060S Graphics
6 NVIDIA-Nemotron-3-Nano-4Bi NVIDIA 4B 143.2 EUR 0.17 5.79M 1.74M AMD Radeon 8060S Graphics
7 North-Mini-Code-1.0i Cohere – 148.0 EUR 0.15 6.74M 2.02M AMD Radeon 8060S Graphics
8 NVIDIA-Nemotron-3-Nano-30B-A3Bi NVIDIA 30B 144.1 EUR 0.17 5.98M 1.79M AMD Radeon 8060S Graphics
9 Qwen-AgentWorld-35B-A3Bi Qwen (Alibaba) 35B 108.6 EUR 0.19 5.30M 1.59M AMD Radeon 8060S Graphics
10 Qwen2.5-32B-Instruct-AWQi Qwen (Alibaba) 32B 155.9 EUR 0.14 7.38M 2.21M NVIDIA RTX PRO 6000 Blackwell Workstation Edition
11 NVIDIA-Nemotron-3-Nano-30B-A3B-BF16i NVIDIA – 236.1 EUR 0.13 7.71M 2.31M NVIDIA RTX PRO 6000 Blackwell Workstation Edition
12 Qwen3-Omni-30B-A3B-Thinkingi Qwen (Alibaba) 30B 342.3 EUR 0.076 13.17M 3.95M NVIDIA RTX PRO 6000 Blackwell Workstation Edition
13 Qwen3.6-27Bi Qwen (Alibaba) 27B 25.7 EUR 0.54 1.84M 552.41K AMD Radeon 8060S Graphics
14 gpt-oss-20bi OpenAI 21B 302.2 EUR 0.043 23.21M 6.96M NVIDIA GeForce RTX 5090
15 Qwen3-30B-A3B-Instruct-2507i Qwen (Alibaba) 30B 251.2 EUR 0.049 20.61M 6.18M NVIDIA GeForce RTX 3090 Ti
16 Qwen3-Coder-30B-A3B-Instructi Qwen (Alibaba) 30B 251.0 EUR 0.062 16.17M 4.85M NVIDIA GeForce RTX 5090
17 MiniMax-M2.5i MiniMax 230B 62.7 EUR 1.66 601.92K 180.58K 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition
18 Qwen3.5-35B-A3Bi Qwen (Alibaba) 35B 230.9 EUR 0.067 14.87M 4.46M NVIDIA RTX PRO 6000 Blackwell Workstation Edition
19 Ministral-3-14B-Reasoning-2512i Mistral AI 14B 236.6 EUR 0.068 14.78M 4.43M NVIDIA GeForce RTX 5090
20 Qwen3-30B-A3B-Thinking-2507i Qwen (Alibaba) 30B 253.2 EUR 0.050 20.20M 6.06M NVIDIA GeForce RTX 3090 Ti
21 Magistral-Small-2509i Mistral AI 24B 218.3 EUR 0.070 14.21M 4.26M NVIDIA GeForce RTX 5090
22 Qwen3.6-35B-A3Bi Qwen (Alibaba) 35B 232.5 EUR 0.070 14.29M 4.29M NVIDIA GeForce RTX 5090
23 Laguna-XS-2.1-INT4i Poolside – 90.0 EUR 0.72 1.39M 417.55K 3× AMD Radeon AI PRO R9700
24 Devstral-Small-2-24B-Instruct-2512i Mistral AI 24B 231.6 EUR 0.065 15.48M 4.64M NVIDIA GeForce RTX 5090
25 Codestral-22B-v0.1i Mistral AI 22B 230.3 EUR 0.063 15.82M 4.75M NVIDIA GeForce RTX 5090
26 OpenReasoning-Nemotron-32Bi NVIDIA 32B 32.0 EUR 0.93 1.08M 322.93K AMD Radeon 8060S Graphics
27 Seed-OSS-36B-Instructi ByteDance Seed – 38.6 EUR 0.46 2.20M 658.57K AMD Radeon 8060S Graphics
28 Devstral-Small-2507i Mistral AI 24B 209.6 EUR 0.062 16.05M 4.81M NVIDIA GeForce RTX 5090
29 gpt-oss-120bi OpenAI 117B 265.7 EUR 0.071 14.18M 4.25M NVIDIA GeForce RTX 5090
30 Qwen3-Coder-Nexti Qwen (Alibaba) – 78.6 EUR 0.71 1.41M 423.20K 3× AMD Radeon AI PRO R9700
31 Mistral-Small-3.1-24B-Instruct-2503i Mistral AI 24B 189.5 EUR 0.071 14.11M 4.23M NVIDIA GeForce RTX 5090
32 gemma-4-E4B-iti Google 4B 51.3 EUR 0.25 4.07M 1.22M AMD Radeon 8060S Graphics
33 GLM-4.5-Airi Z.ai (Zhipu) 106B 257.7 EUR 0.061 16.35M 4.90M NVIDIA GeForce RTX 5090
34 gemma-4-12B-iti Google 12B 48.5 EUR 0.25 4.06M 1.22M AMD Radeon 8060S Graphics
35 gemma-4-E2B-iti Google 2B 49.2 EUR 0.25 4.06M 1.22M AMD Radeon 8060S Graphics
36 Gemma-4-26B-A4B-iti Google 26B 49.1 EUR 0.16 6.10M 1.83M AMD Radeon 8060S Graphics
37 Llama-3_3-Nemotron-Super-49B-v1_5i NVIDIA 49B 21.0 EUR 1.37 728.17K 218.45K AMD Radeon 8060S Graphics
38 gemma-4-31B-iti Google 31B 48.3 EUR 0.24 4.10M 1.23M AMD Radeon 8060S Graphics
39 Qwen2.5-Coder-32B-Instructi Qwen (Alibaba) 32B 224.4 EUR 0.046 21.89M 6.57M NVIDIA GeForce RTX 5090
40 Laguna-S-2.1-INT4i Poolside – 42.0 EUR 1.57 636.18K 190.85K 3× AMD Radeon AI PRO R9700
41 gemma-4-26B-A4B-it-assistanti Google 26B 38.6 EUR 0.26 3.80M 1.14M AMD Radeon 8060S Graphics
42 Nemotron-H-47B-Reasoning-128Ki NVIDIA 47B 12.6 EUR 1.84 544.34K 163.30K AMD Radeon 8060S Graphics
43 Mistral-Small-4-119B-2603i Mistral AI 119B 32.1 EUR 1.89 528.00K 158.40K 3× AMD Radeon AI PRO R9700
44 Step-3.5-Flashi StepFun – 22.6 EUR 3.51 285.11K 85.53K 3× AMD Radeon AI PRO R9700
45 gemma-4-31b-it-int8i Google 31B 14.6 EUR 4.55 219.79K 65.94K 3× AMD Radeon AI PRO R9700
46 Minimax-2.5i MiniMax 230B 13.8 EUR 6.05 165.38K 49.62K 3× AMD Radeon AI PRO R9700
47 Mistral-Medium-3.5-128Bi Mistral AI 128B 13.9 EUR 5.24 190.92K 57.28K 3× AMD Radeon AI PRO R9700
48 DeepSeek-V4-Flash-284B-A13Bi DeepSeek – 15.5 EUR 4.98 200.93K 60.28K 3× AMD Radeon AI PRO R9700
49 command-a-reasoning-08-2025i Cohere – 7.8 EUR 14.19 70.47K 21.14K 3× AMD Radeon AI PRO R9700
50 MiniMax-M2.7i MiniMax 230B 12.2 EUR 6.55 152.57K 45.77K 3× AMD Radeon AI PRO R9700
51 Devstral-2-123B-Instruct-2512i Mistral AI 123B 6.5 EUR 17.05 58.65K 17.59K 3× AMD Radeon AI PRO R9700
52 NVIDIA-Nemotron-3-Super-120B-A12Bi NVIDIA 120B 15.2 EUR 4.76 210.23K 63.07K 3× AMD Radeon AI PRO R9700

Low token costs do not automatically mean high model quality. For quality see the harness benchmarks.