80 GB VRAM and 2,039 GB/s of memory bandwidth. The strongest single-device fits at 4-bit are Gemma 4 31B (~130 tok/s), Qwen3 32B (~124 tok/s), gpt-oss-120b (~715 tok/s), QwQ 32B (~124 tok/s) and gpt-oss-20b (~971 tok/s).
49 open-weight models fit in A100 80 GB at 4-bit, ranked by Analyzer Index. Renting costs $1,095 a month at 24 h/day.
| Model | Index | Memory | Est. tok/s | API $/month | Verdict |
|---|---|---|---|---|---|
| Gemma 4 31B Google31.3B | 57.4 | 19 GB | ~130 | $18.50 | API is cheaper |
| Qwen3 32B Alibaba (Qwen)32.8B | 50.0 | 20 GB | ~124 | $15.00 | API is cheaper |
| gpt-oss-120b OpenAI116.8B · 5.7B active | 45.8 | 70 GB | ~715 | $6.75 | API is cheaper |
| QwQ 32B Alibaba (Qwen)32.8B | 37.2 | 20 GB | ~124 | — | no API price |
| gpt-oss-20b OpenAI20.9B · 4.2B active | 35.2 | 13 GB | ~971 | $4.05 | API is cheaper |
| Gemma 3 27B Google27.4B | 28.5 | 16 GB | ~149 | $12.00 | API is cheaper |
| Llama 3.1 70B Instruct Meta70.6B | 24.3 | 42 GB | ~57.8 | $50.00 | API is cheaper |
| Llama 3.3 70B Instruct Meta70.6B | 20.9 | 42 GB | ~57.8 | $18.00 | API is cheaper |
| Llama 4 Scout Meta108.6B · 17B active | 12.3 | 65 GB | ~240 | $14.00 | API is cheaper |
| Nemotron 3.5 Lightning NVIDIA31.6B · 3B active | — | 19 GB | ~1359 | $5.50 | API is cheaper |
| Qwen3.6 35B A3B Alibaba (Qwen)36B · 3B active | — | 22 GB | ~1359 | $22.50 | API is cheaper |
| Qwen3.5-35B-A3B Alibaba (Qwen)36B · 3B active | — | 22 GB | ~1359 | $25.25 | API is cheaper |
| Nemotron 3 Nano 30B A3B NVIDIA31.6B · 3B active | — | 19 GB | ~1359 | $10.00 | API is cheaper |
| Qwen3 VL 30B A3B Thinking Alibaba (Qwen)31.1B · 3B active | — | 19 GB | ~1359 | $54.00 | API is cheaper |
| Qwen3 VL 30B A3B Instruct Alibaba (Qwen)31.1B · 3B active | — | 19 GB | ~1359 | $30.00 | API is cheaper |
| Qwen3 Next 80B A3B Thinking Alibaba (Qwen)81.3B · 3B active | — | 49 GB | ~1359 | $45.00 | API is cheaper |
| Qwen3 Next 80B A3B Instruct Alibaba (Qwen)81.3B · 3B active | — | 49 GB | ~1359 | $29.20 | API is cheaper |
| Qwen3 Coder 30B A3B Instruct Alibaba (Qwen)30.5B · 3B active | — | 18 GB | ~1359 | $14.00 | API is cheaper |
| Qwen3 30B A3B Instruct 2507 Alibaba (Qwen)30.5B · 3B active | — | 18 GB | ~1359 | $16.50 | API is cheaper |
| Hunyuan A13B Instruct Tencent (Hunyuan)80.4B · 3B active | — | 48 GB | ~1359 | $28.25 | API is cheaper |
| Qwen3 30B A3B Alibaba (Qwen)30.5B · 3B active | — | 18 GB | ~1359 | $24.50 | API is cheaper |
| Ministral 3 3B 2512 Mistral3.8B | — | 2 GB | ~1073 | $12.50 | API is cheaper |
| Qwen3 Coder Next Alibaba (Qwen)79.7B · 3.9B active | — | 48 GB | ~1046 | $32.00 | API is cheaper |
| Gemma 4 26B A4B Google25.8B · 4B active | — | 16 GB | ~1020 | $9.70 | API is cheaper |
| Gemma 3 4B Google4.3B | — | 3 GB | ~948 | $5.25 | API is cheaper |
| GLM 4.7 Flash Zhipu AI (GLM)31.2B · 4.6B active | — | 19 GB | ~887 | $16.05 | API is cheaper |
| Mistral Small 4 Mistral119.4B · 7.1B active | — | 72 GB | ~574 | $30.00 | API is cheaper |
| Stable Diffusion 3.5 Large Stability AI8.1B | — | 5 GB | ~503 | — | no API price |
| Qwen3 VL 8B Instruct Alibaba (Qwen)8.8B | — | 5 GB | ~463 | $43.75 | API is cheaper |
| Ministral 3 8B 2512 Mistral8.9B | — | 5 GB | ~458 | $18.75 | API is cheaper |
| Qwen3.5-9B Alibaba (Qwen)9.7B | — | 6 GB | ~420 | $11.25 | API is cheaper |
| FLUX.1 [dev] Black Forest Labs11.9B | — | 7 GB | ~343 | — | no API price |
| Gemma 3 12B Google12.2B | — | 7 GB | ~334 | $7.00 | API is cheaper |
| Ministral 3 14B 2512 Mistral13.9B | — | 8 GB | ~293 | $25.00 | API is cheaper |
| GLM 4.6V Zhipu AI (GLM)107.7B · 14.3B active | — | 65 GB | ~285 | $52.50 | API is cheaper |
| GLM 4.5V Zhipu AI (GLM)107.7B · 14.3B active | — | 65 GB | ~285 | $105 | API is cheaper |
| Qwen3 14B Alibaba (Qwen)14.8B | — | 9 GB | ~276 | $15.50 | API is cheaper |
| GLM 4.5 Air Zhipu AI (GLM)110.5B · 15B active | — | 66 GB | ~272 | $34.25 | API is cheaper |
| Mistral Small 3 Mistral23.6B | — | 14 GB | ~173 | $7.00 | API is cheaper |
| Mistral Small 3.2 24B Mistral24B | — | 14 GB | ~170 | $12.50 | API is cheaper |
| Mistral Small 3.1 24B Mistral24B | — | 14 GB | ~170 | $48.97 | API is cheaper |
| Voxtral Small 24B 2507 Mistral24.3B | — | 15 GB | ~168 | $17.50 | API is cheaper |
| Qwen3.8 27B Alibaba (Qwen)27.8B | — | 17 GB | ~147 | $14.50 | API is cheaper |
| Qwen3.6 27B Alibaba (Qwen)27.8B | — | 17 GB | ~147 | $70.00 | API is cheaper |
| Qwen3.5-27B Alibaba (Qwen)27.8B | — | 17 GB | ~147 | $58.50 | API is cheaper |
| Muse Glimmer 30B Meta29.8B | — | 18 GB | ~137 | $60.00 | API is cheaper |
| Qwen2.5 72B Instruct Alibaba (Qwen)72.7B | — | 44 GB | ~56.1 | $46.00 | API is cheaper |
| Qwen2.5 VL 72B Instruct Alibaba (Qwen)73.4B | — | 44 GB | ~55.6 | $105 | API is cheaper |
| Command A Cohere111.1B | — | 67 GB | ~36.7 | $500 | API is cheaper |
Memory need is weights × 1.2 for runtime overhead. Speed is a theoretical single-stream ceiling (memory bandwidth ÷ active-weight bytes); real engines usually reach 50–80% of it, and batching many requests raises total throughput well beyond one stream. Rental and purchase prices are typical figures checked 2026-10-14; power and engineering time are not included.