💰
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
Settings are stored in your browser and shared via the URL. Power draw estimated from GPU TDP + CPU (idle + 15 %) + board. Methodology →
Model economics
90 Models
Most economical setup per model. No quality judgement – only measured cost per token. Number after the model = runs · HW configs.
| # | Model | Gen tok/s i | EUR / 1M tok i | Tok/€ | Tok/kWh | cheapest HW |
|---|---|---|---|---|---|---|
| 1 | Kolibri-178.1BiAleph Alpha ↗ Source | 456.6 | EUR 0.024 | 41.78M | 12.53M | NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 2 | Llama-3.1-Nemotron-Nano-4B-v1.14.5BiNVIDIA ↗ Source | 283.4 | EUR 0.073 | 13.79M | 4.14M | AMD Radeon RX 7900 XTX |
| 3 | Nemotron-3-Nano-4B4BiNVIDIA ↗ Source | 596.4 | EUR 0.022 | 46.33M | 13.90M | NVIDIA GeForce RTX 5070 Ti |
| 4 | gemma-4-E4B-it8BiGoogle ↗ Source | 454.2 | EUR 0.024 | 41.90M | 12.57M | NVIDIA GeForce RTX 5070 Ti |
| 5 | NVIDIA-Nemotron-3.5-Super-VL-120B-A12B120BiNVIDIA ↗ Source | 445.0 | EUR 0.016 | 62.49M | 18.75M | NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 6 | gemma-4-26B-A4B-it-assistant26BiGoogle ↗ Source | 58.3 | EUR 0.13 | 7.45M | 2.24M | AMD Radeon 8060S Graphics |
| 7 | Llama-3.2-1B-Instruct1BiMeta ↗ Source | 388.1 | EUR 0.084 | 11.96M | 3.59M | 2× AMD Radeon RX 7900 XTX |
| 8 | Gemma-4-26B-A4B26BiGoogle ↗ Source | 155.4 | EUR 0.12 | 8.54M | 2.56M | NVIDIA RTX A6000 |
| 9 | Gemma-4-26B-A4B-it26BiGoogle ↗ Source | 392.6 | EUR 0.026 | 39.06M | 11.72M | NVIDIA GeForce RTX 5070 Ti |
| 10 | Ornith-1.5-35B-A3B35BiDeepReinforce ↗ Source | 305.9 | EUR 0.089 | 11.20M | 3.36M | 2× NVIDIA RTX A6000 |
| 11 | Qwen3-0.6B0.6BiQwen (Alibaba) ↗ Source | 309.5 | EUR 0.045 | 22.46M | 6.74M | AMD Radeon AI PRO R9700 |
| 12 | Ornith-1.0-9B9BiDeepReinforce ↗ Source | 339.0 | EUR 0.023 | 44.03M | 13.21M | NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 13 | Llama-3.2-3B-Instruct3BiMeta ↗ Source | 218.5 | EUR 0.063 | 15.89M | 4.77M | AMD Radeon AI PRO R9700 |
| 14 | Qwen3-1.7B1.7BiQwen (Alibaba) ↗ Source | 269.1 | EUR 0.051 | 19.71M | 5.91M | AMD Radeon AI PRO R9700 |
| 15 | Qwen2.5-3B-Instruct3BiQwen (Alibaba) ↗ Source | 208.7 | EUR 0.069 | 14.57M | 4.37M | AMD Radeon AI PRO R9700 |
| 16 | Meta-Llama-3.1-8B-Instruct8BiMeta ↗ Source | 324.6 | EUR 0.028 | 35.36M | 10.61M | NVIDIA GeForce RTX 5070 Ti |
| 17 | Laguna-XS-2.1iPoolside ↗ Source | 590.7 | EUR 0.029 | 34.08M | 10.22M | NVIDIA GeForce RTX 5090 |
| 18 | Nemotron-Cascade-2-30B-A3B30BiNVIDIA ↗ Source | 400.5 | EUR 0.029 | 34.99M | 10.50M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 19 | Ornith-1.5-9B9BiDeepReinforce ↗ Source | 223.1 | EUR 0.12 | 8.44M | 2.53M | 2× NVIDIA RTX A6000 |
| 20 | Huihui-Qwen3-VL-30B-A3B-Instruct-abliterated30BiQwen (Alibaba) ↗ Source | 158.2 | EUR 0.11 | 9.34M | 2.80M | AMD Radeon AI PRO R9700 |
| 21 | gemma-4-12B-it12BiGoogle ↗ Source | 265.2 | EUR 0.029 | 34.46M | 10.34M | NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 22 | Qwen2.5-7B-Instruct7BiQwen (Alibaba) ↗ Source | 179.7 | EUR 0.100 | 10.02M | 3.01M | AMD Radeon AI PRO R9700 |
| 23 | gemma-4-E2B-it5BiGoogle ↗ Source | 510.4 | EUR 0.013 | 77.43M | 23.23M | NVIDIA GB10 (DGX Spark) |
| 24 | SmolLM2-1.7B-Instruct1.7BiHuggingFaceTB ↗ Source | 314.2 | EUR 0.11 | 8.86M | 2.66M | 2× AMD Radeon RX 7900 XTX |
| 25 | Qwen3-4B4BiQwen (Alibaba) ↗ Source | 185.1 | EUR 0.077 | 13.01M | 3.90M | AMD Radeon AI PRO R9700 |
| 26 | gemma-3n-E4B-it8BiGoogle ↗ Source | 161.5 | EUR 0.11 | 8.98M | 2.69M | AMD Radeon RX 7900 XTX |
| 27 | Nemotron-3-Nano-Omni-30B-A3B-Reasoning30BiNVIDIA ↗ Source | 432.5 | EUR 0.029 | 34.79M | 10.44M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 28 | North-Mini-Code-1.0iCohere ↗ Source | 396.6 | EUR 0.032 | 30.84M | 9.25M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 29 | Ministral-3-14B-Reasoning-251214BiMistral AI ↗ Source | 243.8 | EUR 0.044 | 22.84M | 6.85M | NVIDIA GeForce RTX 5070 Ti |
| 30 | Qwen3-Omni-30B-A3B-Thinking30BiQwen (Alibaba) ↗ Source | 396.3 | EUR 0.031 | 32.31M | 9.69M | NVIDIA GeForce RTX 5090 |
| 31 | gpt-oss-20b20BiOpenAI ↗ Source | 450.2 | EUR 0.021 | 46.85M | 14.05M | NVIDIA GeForce RTX 5070 Ti |
| 32 | Qwen2.5-Coder-7B-InstructiQwen (Alibaba) ↗ Source | 162.1 | EUR 0.24 | 4.11M | 1.23M | 2× AMD Radeon RX 7900 XTX |
| 33 | Qwen3.6-35B-A3B35BiQwen (Alibaba) ↗ Source | 261.1 | EUR 0.034 | 29.40M | 8.82M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 34 | Qwen3-Coder-30B-A3B-Instruct30BiQwen (Alibaba) ↗ Source | 332.4 | EUR 0.030 | 33.14M | 9.94M | NVIDIA GeForce RTX 5090 |
| 35 | Qwen3-8B8.2BiQwen (Alibaba) ↗ Source | 152.9 | EUR 0.19 | 5.25M | 1.58M | 2× AMD Radeon RX 7900 XTX |
| 36 | Qwen3-30B-A3B30BiQwen (Alibaba) ↗ Source | 145.1 | EUR 0.27 | 3.64M | 1.09M | 2× AMD Radeon RX 7900 XTX |
| 37 | gemma-2-2b-it2BiGoogle ↗ Source | 199.2 | EUR 0.19 | 5.38M | 1.61M | 2× AMD Radeon RX 7900 XTX |
| 38 | Qwen3-30B-A3B-Thinking-250730BiQwen (Alibaba) ↗ Source | 347.3 | EUR 0.031 | 32.51M | 9.75M | NVIDIA GeForce RTX 5090 |
| 39 | Qwen-AgentWorld-35B-A3B35BiQwen (Alibaba) ↗ Source | 342.1 | EUR 0.035 | 28.22M | 8.47M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 40 | Qwen3-30B-A3B-Instruct-250730BiQwen (Alibaba) ↗ Source | 327.4 | EUR 0.031 | 32.41M | 9.72M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 41 | Qwen3-VL-30B-A3B-Instruct30BiQwen (Alibaba) ↗ Source | 331.5 | EUR 0.031 | 31.97M | 9.59M | NVIDIA GeForce RTX 5090 |
| 42 | glm-4.7-flashiZ.ai (Zhipu) ↗ Source | 141.0 | EUR 0.11 | 9.31M | 2.79M | NVIDIA GeForce RTX 3090 Ti |
| 43 | Phi-3.5-mini-instruct3.8Bimicrosoft ↗ Source | 179.2 | EUR 0.20 | 5.03M | 1.51M | 2× AMD Radeon RX 7900 XTX |
| 44 | Nemotron-3-Nano-30B-A3B30BiNVIDIA ↗ Source | 371.0 | EUR 0.029 | 34.82M | 10.44M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 45 | Ornith-1.0-35B35BiDeepReinforce ↗ Source | 485.4 | EUR 0.032 | 31.10M | 9.33M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 46 | Qwen3.5-35B-A3B35BiQwen (Alibaba) ↗ Source | 330.4 | EUR 0.035 | 28.94M | 8.68M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 47 | Nemotron-3.5-Lightning-30B-A3B30BiNVIDIA ↗ Source | 280.7 | EUR 0.017 | 60.37M | 18.11M | NVIDIA GeForce RTX 3090 Ti |
| 48 | Qwen3.8-27B-Uncensored27.00BiQwen (Alibaba) ↗ Source | 90.9 | EUR 0.54 | 1.85M | 555.43K | 2× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 49 | gemma-2-9b-it9BiGoogle ↗ Source | 113.3 | EUR 0.36 | 2.75M | 824.86K | 2× AMD Radeon RX 7900 XTX |
| 50 | Qwen3-14B14BiQwen (Alibaba) ↗ Source | 122.9 | EUR 0.21 | 4.68M | 1.40M | 2× AMD Radeon RX 7900 XTX |
| 51 | Qwen3.8-Flash-Next180BiQwen (Alibaba) ↗ Source | 243.4 | EUR 0.016 | 62.58M | 18.78M | NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 52 | Devstral-Small-2-24B-Instruct-251224BiMistral AI ↗ Source | 155.5 | EUR 0.072 | 13.91M | 4.17M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 53 | Devstral-Small-250724BiMistral AI ↗ Source | 150.8 | EUR 0.079 | 12.69M | 3.81M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 54 | Codestral-22B-v0.122BiMistral AI ↗ Source | 130.5 | EUR 0.081 | 12.38M | 3.72M | NVIDIA GeForce RTX 5090 |
| 55 | Magistral-Small-250924BiMistral AI ↗ Source | 171.7 | EUR 0.070 | 14.28M | 4.28M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 56 | Qwen3.6-27B27BiQwen (Alibaba) ↗ Source | 130.7 | EUR 0.097 | 10.28M | 3.08M | NVIDIA GeForce RTX 5090 |
| 57 | Mistral-Small-24B-Instruct-2501iMistral AI ↗ Source | 92.8 | EUR 0.41 | 2.45M | 736.20K | 2× AMD Radeon RX 7900 XTX |
| 58 | Qwen3-Coder-NextiQwen (Alibaba) ↗ Source | 282.9 | EUR 0.043 | 23.36M | 7.01M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 59 | Ling-3.0-flash127.5BiinclusionAI ↗ Source | 95.4 | EUR 0.046 | 21.74M | 6.52M | NVIDIA GB10 (DGX Spark) |
| 60 | gemma-2-27b-it27BiGoogle ↗ Source | 69.4 | EUR 0.55 | 1.81M | 543.21K | 2× AMD Radeon RX 7900 XTX |
| 61 | Qwen3-32B32BiQwen (Alibaba) ↗ Source | 70.0 | EUR 0.53 | 1.88M | 563.64K | 2× AMD Radeon RX 7900 XTX |
| 62 | Laguna-S-2.1-INT4iPoolside ↗ Source | 287.6 | EUR 0.064 | 15.57M | 4.67M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 63 | gpt-oss-120b120BiOpenAI ↗ Source | 291.7 | EUR 0.040 | 24.90M | 7.47M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 64 | gemma-4-31B-it31BiGoogle ↗ Source | 78.6 | EUR 0.22 | 4.64M | 1.39M | NVIDIA GeForce RTX 3090 Ti |
| 65 | Mistral-Small-3.1-24B-Instruct-250324BiMistral AI ↗ Source | 67.0 | EUR 0.17 | 5.92M | 1.77M | AMD Radeon PRO W7900 Dual Slot |
| 66 | Qwen3.8-27B27BiQwen (Alibaba) ↗ Source | 101.0 | EUR 0.13 | 7.63M | 2.29M | 2× NVIDIA RTX A6000 |
| 67 | OpenReasoning-Nemotron-32B32BiNVIDIA ↗ Source | 109.9 | EUR 0.11 | 9.29M | 2.79M | NVIDIA GeForce RTX 5090 |
| 68 | Muse-Glimmer-30B30BiMeta ↗ Source | 85.2 | EUR 0.15 | 6.87M | 2.06M | AMD Radeon AI PRO R9700 |
| 69 | Mistral-Small-4-119B-2603119BiMistral AI ↗ Source | 93.1 | EUR 0.056 | 17.89M | 5.37M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 70 | Qwen2.5-32B-Instruct-AWQ32BiQwen (Alibaba) ↗ Source | 127.7 | EUR 0.11 | 9.51M | 2.85M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 71 | Seed-OSS-36B-Instruct36BiByteDance Seed ↗ Source | 104.1 | EUR 0.062 | 16.21M | 4.86M | NVIDIA GeForce RTX 5090 |
| 72 | minimax-m2-awq230BiMiniMax ↗ Source | 114.8 | EUR 0.15 | 6.69M | 2.01M | 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 73 | Qwen2.5-Coder-32B-Instruct32BiQwen (Alibaba) ↗ Source | 95.9 | EUR 0.11 | 9.33M | 2.80M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 74 | Qwen3.5-122B-A10B122BiQwen (Alibaba) ↗ Source | 86.9 | EUR 0.15 | 6.90M | 2.07M | 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 75 | Step-3.5-FlashiStepFun ↗ Source | 14.5 | EUR 3.58 | 279.67K | 83.90K | 2× AMD Radeon RX 7900 XTX |
| 76 | Llama-3.3-Nemotron-Super-49B-v1.549BiNVIDIA ↗ Source | 75.4 | EUR 0.15 | 6.79M | 2.04M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 77 | GLM-4.5-Air106BiZ.ai (Zhipu) ↗ Source | 120.2 | EUR 0.081 | 12.34M | 3.70M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 78 | Nemotron-H-47B-Reasoning-128K47BiNVIDIA ↗ Source | 45.1 | EUR 0.26 | 3.81M | 1.14M | NVIDIA GeForce RTX 5090 |
| 79 | Nemotron-3-Super-120B-A12B120BiNVIDIA ↗ Source | 81.4 | EUR 0.089 | 11.20M | 3.36M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 80 | DeepSeek-V4-Flash290.9BiDeepSeek ↗ Source | 15.4 | EUR 5.14 | 194.49K | 58.35K | 4× AMD Radeon AI PRO R9700 |
| 81 | DeepSeek-V4-Flash-284B-A13B284BiDeepSeek ↗ Source | 65.6 | EUR 0.21 | 4.71M | 1.41M | 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 82 | MiniMax-M2.5230BiMiniMax ↗ Source | 84.1 | EUR 0.11 | 9.05M | 2.71M | NVIDIA GB10 (DGX Spark) |
| 83 | Laguna-M.1iPoolside ↗ Source | 39.7 | EUR 0.24 | 4.12M | 1.24M | 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 84 | Qwen2.5-72B-Instruct72BiQwen (Alibaba) ↗ Source | 45.2 | EUR 0.22 | 4.46M | 1.34M | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 85 | command-a-reasoning-08-2025111BiCohere ↗ Source | 38.1 | EUR 0.34 | 2.93M | 877.76K | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 86 | Mistral-Medium-3.5-128B128BiMistral AI ↗ Source | 9.4 | EUR 0.38 | 2.64M | 791.69K | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 87 | MiniMax-M2.7230BiMiniMax ↗ Source | 64.4 | EUR 0.067 | 14.93M | 4.48M | NVIDIA GB10 (DGX Spark) |
| 88 | Llama-3.1-Nemotron-Ultra-253B-v1253BiNVIDIA ↗ Source | 4.3 | EUR 1.32 | 758.23K | 227.47K | 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition |
| 89 | Devstral-2-123B-Instruct-2512123BiMistral AI ↗ Source | 20.9 | EUR 0.38 | 2.67M | 799.97K | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
| 90 | glm-5.2-colibri744BiZ.ai (Zhipu) ↗ Source | 0.0 | EUR 1,788.19 | 559 | 168 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition |
Low token costs do not automatically mean high model quality. For quality see the harness benchmarks.