NVIDIA · consumer
Titan Xp
Titan Xp has 12 GB of VRAM at 548 GB/s — about 11.16 GiB usable after driver and compositor overhead. 656 of 2118 indexed models fit at 128K context with f16 KV.
Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
12 GB
GDDR5X
Bandwidth
548 GB/s
384-bit bus
Tensor FP16
—
dense
TDP
250 W
$1200 MSRP
text 499vision language 78video 14embedding 15audio tts 18audio asr 31image 1
What fits at 128K context
largest quantization that fits, per model · 656 of 2118 indexed
| Model | Best quant | Params | Weights● | KV● | Total in memory◐ | Headroom◐ | tok/s◐ |
|---|---|---|---|---|---|---|---|
| HuatuoGPT-o1-7B | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| DeepHat-V1-7B | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| openhands-lm-7b-v0.1 | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Qwen2.5-Coder-7B-Instruct-abliterated | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Qwen2.5-Coder-7B-Instruct | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| UwU-7B-Instruct | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Qwen2.5-Math-7B-Instruct | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Qwen2.5-7B-Instruct | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| OREAL-DeepSeek-R1-Distill-Qwen-7B | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Qwen2.5-7B-Instruct-1M | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| DeepSeek-R1-Distill-Qwen-7B | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Qwen2-VL-7B-Instruct-abliterated | Q2_K_L | 8.3B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| olmOCR-2-7B-1025 | Q2_K_L | 8.3B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Qwen2-VL-7B-Instruct | Q2_K_L | 8.3B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Qwen2.5-VL-7B-Instruct | Q2_K_L | 8.3B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Fara-7B | Q2_K_L | 8.3B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| EVA-Qwen2.5-7B-v0.1 | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Human-Like-Qwen2.5-7B-Instruct | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| UI-TARS-7B-DPO | Q2_K_L | 8.3B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| zeta | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Hercules-5.0-Qwen2-7B | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Wan2.1-FLF2V-14B-720P | Q4_1 | 16.4B | 10.32 GiB | 0.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| MiniCPM-o-2_6 | Q2_K_L | 8.7B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Arcee-Maestro-7B-Preview | Q2_K_L | 7.6B | 3.30 GiB | 7.00 GiB | 11.16 GiB | 0.00 GiB | 38±12.9% |
| Wan2.1-I2V-14B-480P | Q4_1 | 16.4B | 10.32 GiB | 0.00 GiB | 11.15 GiB | 0.01 GiB | 38±12.9% |
| Wan2.1-I2V-14B-720P | Q4_1 | 16.4B | 10.32 GiB | 0.00 GiB | 11.15 GiB | 0.01 GiB | 38±12.9% |
| SmolLM3-3B | Q3_K_S | 3.1B | 1.33 GiB | 9.00 GiB | 11.15 GiB | 0.01 GiB | 38±12.9% |
| LocateAnything-3B | Q8_0 | 3.8B | 5.83 GiB | 4.50 GiB | 11.14 GiB | 0.02 GiB | 38±12.9% |
| GrammarCoder-7B-Base | I1-IQ3_S | 7.6B | 3.27 GiB | 7.00 GiB | 11.12 GiB | 0.04 GiB | 38±12.9% |
| DeepHat-V1-7B-Heretic-Abliterated | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| ShizhenGPT-7B-VL | I1-IQ3_S | 8.3B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| MathSmith-DS-Qwen-7B-LongCoT | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| AstraGPTCoder-7B | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen2.5-Coder-7B-Instruct-Ghidra-v2 | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| EsDrac-v1-7B | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Hemlock-Apothecary-7B-GRPO-e3 | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Hemlock2-Coder-7B-GRPO | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| shellwhiz-7b | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen2.5-Coder-7B-Instruct-OBLITERATED-advanced | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen-STEM-Specialist-7B | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| VulnLLM-R-7B | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Garnet-OCR-7B-0422 | I1-IQ3_S | 8.3B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Video-R1-7B | I1-IQ3_S | 8.3B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| HARC-Qwen2.5-7B-Instruct | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen2.5-Coder-7B-Abliterated | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Bozdogan-7B | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen2.5-7B-Instruct-abliterated-v2 | IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Crazy-AI-Model | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| turbo-ai-7b | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| DeepSeek-R1-Distill-Qwen-7B-abliterated-v2 | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Ghosty-7B | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| SP-7B | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen2.5-Coder-7B-Instruct-Uncensored | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen2.5-VL-7B-Instruct-abliterated | I1-IQ3_S | 8.3B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| DeepSeek-R1-Distill-Qwen-8B-Abliterated | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen2.5-7B | IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen2.5-VL-7B-Instruct-heretic | I1-IQ3_S | 8.3B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| AWARES-Qwen2.5-VL-7B | I1-IQ3_S | 8.3B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
| Qwen2.5-7B-Instruct-Uncensored | I1-IQ3_S | 7.6B | 3.26 GiB | 7.00 GiB | 11.11 GiB | 0.05 GiB | 38±12.9% |
Speed is modeled, not measured: decode is memory-bandwidth bound, so tokens per second is bytes read per token against achievable bandwidth. Mixture-of-experts models carry a wider band because only the routed experts are read each step, and few have been measured publicly.
Questions people ask
- What AI models can a Titan Xp run?
- 656 of 2118 indexed open-weight models fit a Titan Xp at 131,072 context with f16 KV cache, the largest being HuatuoGPT-o1-7B at Q2_K_L. That covers text, vision-language, image, video and speech models.
- How much usable memory does a Titan Xp actually have?
- Its nameplate is 12 GB, but about 11.16 GiB is available to a model once driver and compositor overhead is accounted for.
- Is a Titan Xp fast for local AI?
- Its memory bandwidth is 548 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.