RentGPU RentGPU
寻找 GPU →
AI models

Every way
to use the major models.

Closed models like Claude and GPT — link to the cheapest API provider. Open-weights like Llama, Kimi, DeepSeek — choose hosted inference or self-host on rented GPUs.

10 tracked · 10 open weights · 0 closed APIs
Modality All Text Multimodal Code Image Audio Video Vision Embedding
License All Open weights Closed / API
Size All ≤8B (edge) 8–30B (laptop) 30–100B (workstation) 100B+ (datacenter)

10 models match — reset filters

Open-weights models.

Run yourself on cheap GPUs, or use a hosted-inference provider.

Qwen 2.5 Coder 32B

33B
by Alibaba (Qwen Team) · Qwen · 128,000 ctx

Alibaba's open-weight coding model — best in class for 32B.

Code Llama 70B

70B
by Meta AI · Code Llama · 16,384 ctx

Meta's largest code-specialised Llama.

Code Llama 34B

34B
by Meta AI · Code Llama · 16,384 ctx

Code Llama 13B

13B
by Meta AI · Code Llama · 16,384 ctx

Code Llama 7B

7B
by Meta AI · Code Llama · 16,384 ctx

DeepSeek Coder V2 236B

236B
by DeepSeek · DeepSeek Coder · 128,000 ctx

DeepSeek's MoE coding model — 236B total, 21B active.

DeepSeek Coder V2 Lite

16B
by DeepSeek · DeepSeek Coder · 128,000 ctx

16B MoE / 2.4B active — laptop-class coder.

DeepSeek Coder 33B

33B
by DeepSeek · DeepSeek Coder · 16,384 ctx

DeepSeek Coder 6.7B

7B
by DeepSeek · DeepSeek Coder · 16,384 ctx

IBM Granite Code 8B

8B
by IBM Research · Granite · 4,096 ctx