- Services
Every open model in our directory, ranked by how well it runs on NVIDIA RTX A2000 (6GB) at 4-bit. Grades weigh whether the model fits in memory and how fast it should run.
NVIDIA RTX A2000 has 6.0 GB of VRAM, enough to run 155 of the 344 open models we track at 4-bit. The largest that still fits well is PersonaPlex 7B (7B), and the top-rated is PersonaPlex 7B at about 48 tok/s.
155 of 344 models run comfortably on NVIDIA RTX A2000.
See full NVIDIA RTX A2000 specs and pricingThe top models this device can run at 4-bit, ranked by fit and speed.
| Model | Grade | Speed | VRAM |
|---|---|---|---|
| PersonaPlex 7BNVIDIA | SS | 48.4 tok/s | 4.8 GB |
| Llama 2 7B ChatMeta | SS | 48.4 tok/s | 4.8 GB |
| MiniCPM-V 4.7 1Bopenbmb | SS | 49.9 tok/s | 4.6 GB |
| Gemma 4 E2B ITGoogle | SS | 62.5 tok/s | 3.7 GB |
| MiniCPM5-2Bopenbmb | SS | 50.4 tok/s | 4.6 GB |
| VibeThinker-3BWeiboAI | SS | 60.8 tok/s | 3.8 GB |
| ALoDLM-8BAmazon | AA | 42.9 tok/s | 5.4 GB |
| Apertus 8BEPFL, ETH Zurich, CSCS | AA | 42.9 tok/s | 5.4 GB |
| Llama 3 8B InstructMeta | AA | 40.9 tok/s | 5.7 GB |
| LFM2.5-2.6BLiquid AI | AA | 47.5 tok/s | 4.9 GB |
| MiniMax M3.1 FlashMiniMax | AA | 463.7 tok/s | 0.5 GB |
| Jais 3MBZUAI and Inception42 | AA | 463.7 tok/s | 0.5 GB |
PersonaPlex 7BNVIDIA | 7B | SS | 48.4 tok/s | 4.8 GB | |
Llama 2 7B ChatMeta | 7B | SS | 48.4 tok/s | 4.8 GB | |
MiniCPM-V 4.7 1Bopenbmb | 1.55B | SS | 49.9 tok/s | 4.6 GB | |
Gemma 4 E2B ITGoogle | 2B | SS | 62.5 tok/s | 3.7 GB | |
MiniCPM5-2Bopenbmb | 2.52B | SS | 50.4 tok/s | 4.6 GB | |
VibeThinker-3BWeiboAI | 3B | SS | 60.8 tok/s | 3.8 GB | |
ALoDLM-8BAmazon | 8B | AA | 42.9 tok/s | 5.4 GB | |
Apertus 8BEPFL, ETH Zurich, CSCS | 8B | AA | 42.9 tok/s | 5.4 GB | |
Ad Tools for AI BuildersAll the hardware and tools you need to run private AI at your office or at home.Shop Now | |||||
| 8B | AA | 40.9 tok/s | 5.7 GB | ||
LFM2.5-2.6BLiquid AI | 2.69B | AA | 47.5 tok/s | 4.9 GB | |
MiniMax M3.1 FlashMiniMax | 0B | AA | 463.7 tok/s | 0.5 GB | |
Jais 3MBZUAI and Inception42 | 0B | AA | 463.7 tok/s | 0.5 GB | |
BolmoAi2 | 0B | AA | 463.7 tok/s | 0.5 GB | |
LensVLM-9BApple | 9B | BB | 38.5 tok/s | 6.0 GB | |
| 9B | BB | 38.5 tok/s | 6.0 GB | ||
Youtu-Parsing-Omnitencent | 5.33B | BB | 36.3 tok/s | 6.4 GB | |
Ad Vast.aiAffordable on-demand GPU rentals for training and inference. Pick from thousands of hosts.Rent a GPU | |||||
Mistral 7B InstructMistral AI | 7B | BB | 36.3 tok/s | 6.4 GB | |
Qwen3.8-MaxAlibaba | 2.4B(95B active) | FF | 0.3 tok/s | 815.6 GB | |
| 3.98B | FF | 20.8 tok/s | 11.2 GB | ||
| 4B | FF | 20.7 tok/s | 11.2 GB | ||
Gemma 4 E4B ITGoogle | 4B | FF | 33.5 tok/s | 6.9 GB | |
Gemma 3 4B ITGoogle | 4B | FF | 33.5 tok/s | 6.9 GB | |
K2-Horizon-7BInstitute of Foundation Models (IFM) | 7B | FF | 6.8 tok/s | 33.9 GB | |
| 8B | FF | 17.4 tok/s | 13.3 GB | ||
| AdTools for AI BuildersAll the hardware and tools you need to run private AI at your office or at home.Shop Now | |||||
LFM2.5-8B-A1BLiquid AI | 8.3B(1.5B active) | FF | 79.8 tok/s | 6.8 GB | |
Model Compatibility
Use the calculator to weigh this device against any other, model by model, with speed and fit scores.
Straight answers to the questions we hear most often.
Can't find what you're looking for? Book a Discovery Call
PersonaPlex 7B is the highest-rated model that runs on NVIDIA RTX A2000 at 4-bit, at about 48 tok/s. In total, NVIDIA RTX A2000 can run 155 of the 344 open models we track.
We size each model at 4-bit (Q4_K_M), add the KV cache for its context length and a fixed runtime overhead, then compare that to this device VRAM. Each model is graded on how comfortably it fits and how fast it should run on this card.
It is an estimate of decode speed in tokens per second at 4-bit, based on this device memory bandwidth and each model size. Real speed depends on your runtime, batch size, and context length, so use it to compare models, not as a hard promise.
Lower-bit quantization shrinks a model so more of them fit. These grades already assume 4-bit, which is the common local default. Going below 4-bit can squeeze a larger model in at some quality cost; going above needs more VRAM and may not fit.
Mixture-of-experts models only activate a fraction of their parameters at a time, so their runtime memory is closer to the active size than the headline size. We score fit on the active parameters, which is why some very large models still fit.