Titan Xp 12GB

NVIDIA · 12 GB GDDR5X · 548 GB/s · 250 W · MSRP $1,200

The Titan Xp 12GB is a NVIDIA accelerator with 12 GB GDDR5X and about 548 GB/s peak memory bandwidth. AI Hashrate estimates local LLM decode (batch 1) at 8K context: fit counts weights + KV cache + 1 GB runtime overhead — not weights alone. At Q4, about 18 curated models fit fully on this card; at FP16, about 7 fit. Top Q4 speeds: MiniCPM5-1B ≈ 322.9 tok/s (estimated, fits); Qwen3.5-2B ≈ 174.4 tok/s (estimated, fits); DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx ≈ 151.6 tok/s (estimated, fits); Llama-3.2-3B-Instruct ≈ 109.0 tok/s (estimated, fits); Qwen3-4B-Instruct-2507 ≈ 87.2 tok/s (estimated, fits). Relative ranking is more reliable than absolute tok/s. See methodology for the bandwidth formula and measured-anchor policy. Methodology.

Table: decode speed estimates (Q4 / FP16). Measured rows override estimates. MSRP is list price, not live retail. Context default 8K.

Fit checks

• Qwen3-8B — fits · ≈42.5 tok/s Q4 est.
• gemma-4-26B-A4B-it — does not fit · ≈154.1 tok/s Q4 est.
• Qwen2.5-7B-Instruct — fits · ≈45.9 tok/s Q4 est.
• Qwen3.5-9B — fits · ≈39.0 tok/s Q4 est.
• Llama-3.1-8B-Instruct — fits · ≈43.6 tok/s Q4 est.
• gpt-oss-20b — does not fit · ≈216.2 tok/s Q4 est.
• Qwen3.6-35B-A3B — does not fit · ≈109.2 tok/s Q4 est.
• Qwen3.5-35B-A3B — does not fit · ≈117.2 tok/s Q4 est.
• gemma-4-E2B-it — fits · ≈68.4 tok/s Q4 est.
• Qwen3.6-27B — does not fit · ≈17.8 tok/s Q4 est.
• Mistral-7B-Instruct-v0.3 — fits · ≈48.4 tok/s Q4 est.

All 69 model fit checks — see the linked Fits? column in the table above