Local Models on a GPU: Testing Qwen and Gemma on the Tesla V100
The Tesla V100, released in 2017, can run modern LLMs like Qwen35B and GPT-OSS-20B. Generation speed ranges from 38 to 109 tokens per second, depending on the model. The tests covered building websites, creating games and processing documents. The card needs at least 16 GB of memory, but 24 GB is preferable for general-purpose use.
AI-processed from Habr AI; edited by Hamidun News
You can run local AI models on GPUs released as far back as 2017. We tested five popular open models (from Qwen to Gemma) on a Tesla V100 across five real-world scenarios, from text generation to writing code and building games.
Which text models run on the V100?
The test used five open models, including Qwen and Gemma. Speeds range from 38 to 109 tokens per second, depending on the model and quantization.
Can you use an old V100 for local neural networks?
Yes. According to the test, the V100 handles modern LLMs for text generation, coding and other real-world scenarios.
How many tokens per second does the
Tesla V100 deliver with local models? 38–109 tokens per second, depending on the model, its size and quantization.
- Nov 1, 2026MoneyGate safety layer blocked all 42 fraudulent requests to local AI models
- Nov 1, 2026Hugging Face's BERT Tokenizer Gets a 50x Speedup by Skipping Needless Work on ASCII
- Nov 1, 2026ThinkJev Makes AI Think Only About Hard Questions: 28% Fewer Tokens
- Oct 31, 2026Open-source LLM gateway KTMS 1.0 plugs your own GPUs into the cloud and survives outages
Key Questions Answered
Which text models run on the V100?
The test used five open models, including Qwen and Gemma. Speeds range from 38 to 109 tokens per second, depending on the model and quantization.
Can you use an old V100 for local neural networks?
Yes. According to the test, the V100 handles modern LLMs for text generation, coding and other real-world scenarios.
How many tokens per second does the Tesla V100 deliver with local models?
38–109 tokens per second, depending on the model, its size and quantization.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.