llmrun
GoogleGemma

T5gemma S S Prefixlm — Hardware Requirements & GPU Compatibility

Chat

T5gemma S S Prefixlm is a 313M-parameter open language model from Google in the Gemma family. At BF16 it needs about 0.69 GB of VRAM — see which GPUs and Macs can run it below.

313M params · denseGemma Terms47.4K downloads5 likes
S

Runs great on a 24 GB GPU

Runs from8 GB BF16 · 32KBest on24 GB BF16 · ~950 tok/s · 32KMacs from8 GB BF16 · ~104 tok/s

Gemma Terms

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run T5gemma S S Prefixlm

One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.

Recommended

vLLM
one command · needs vLLM installed
$vllm serve google/t5gemma-s-s-prefixlm --max-model-len 16384

vLLM repo with its own quantisation

Download: Weights on Hugging Face

install, OS, other quants

Ways to runOther ways

each opens step-by-step

Run in cloud

Runs on RTX 3060 12GB (10 GB headroom) · BF16

Generation speed
~339 tok/s
generation speed
Cost per 1M output tokens
$0.052
per 1M output tokens

How Much VRAM Does T5gemma S S Prefixlm Need?

Select a quantization level; the listed hardware adheres to it.

QuantizationVRAM+32KOn RTX 4090 (reference)
0.7 GB2.9 GB16-bit · bfloat — Brain floating point 16 — preferred for trainingFits: 32K context · ~950 tok/s
est.calculated, no published file yet

Selected BF16

Brain floating point 16

1.7GBat 16K context

Weights
0.2 GB
KV cache
1.2 GB
Overhead
0.3 GB

Fits RTX 4090 (reference) · 22.3 GB free · ~950 tok/s

≥50 tok/s needs ≥53 GB/s · RTX 5090 ~1688, M5 Ultra ~1246

Which Hardware Can Run T5gemma S S Prefixlm?

BF16 · 0.7 GB

T5gemma S S Prefixlm (BF16) requires 0.7 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 1+ GB is recommended. 114 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 185 devices with unified memory can run T5gemma S S Prefixlm, including Mac Studio M5 Ultra (256 GB), Mac Studio M5 Ultra (512 GB).

By VRAM tier · typical GPUs and Macs at BF16
8 GB · RTX 4060BF16 · 1.7 GB · ~256 tok/s · 32K8 GB tier →
12 GB · RTX 3060 12GBBF16 · 1.7 GB · ~339 tok/s · 32K12 GB tier →
16 GB · RTX 5060 Ti 16GBBF16 · 1.7 GB · ~422 tok/s · 32K16 GB tier →
24 GB · RTX 4090referenceBF16 · 1.7 GB · ~950 tok/s · 32K24 GB tier →
Mac 32 GB · Mac Mini M4BF16 · 1.7 GB · ~122 tok/s · 32KMacs →
Mac 64 GB · Mac Studio M4 MaxBF16 · 1.7 GB · ~554 tok/s · 32KMacs →

Alternatives to T5gemma S S Prefixlm

Gemma family · 17 models →

Similar models

3
ModelScoreVRAMtok/sWhy it is similar
MiniCPM5 2B DSparkOpenBMB—0.6 GB~1310same size
LFM2.5 2.6B DSparkLiquid AI—0.7 GB~1260same size
Lite Oute 1 300MOuteAI—0.6 GB~1191same size
Fine-tunes & variants(1)

Specifications

Publisher
Google
Family
Gemma
Parameters
313M
Release Date
2025-06-19
License
Gemma Terms

Frequently Asked Questions

How much VRAM does T5gemma S S Prefixlm need?

T5gemma S S Prefixlm requires 0.7 GB of VRAM at BF16.

VRAM = Weights + KV Cache + Overhead

Weights = 313M × 16 bits ÷ 8 = 0.6 GB

KV Cache + Overhead ≈ 0.1 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

VRAM usage by quantization

0.7 GB

Learn more about VRAM estimation →

Can I run T5gemma S S Prefixlm on a Mac?

Yes — MacBook Air 13" M3 (8 GB) and 158 other Macs can run T5gemma S S Prefixlm. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 0.7 GB of usable unified memory (RAM minus macOS overhead).

Can I run T5gemma S S Prefixlm locally?

Yes — T5gemma S S Prefixlm can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is T5gemma S S Prefixlm?

At BF16, T5gemma S S Prefixlm can reach ~6957 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~950 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = (bandwidth GB/s ÷ model GB) × efficiency

Example: NVIDIA B200 → 8000 ÷ 0.69 × 0.65 = ~7536 tok/s

Estimated speed at BF16 (0.7 GB)

~7536 tok/s
~950 tok/s
~7536 tok/s
~6957 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of T5gemma S S Prefixlm?

At BF16, the download is about 0.63 GB.

Which GPUs can run T5gemma S S Prefixlm?

114 consumer GPUs can run T5gemma S S Prefixlm at BF16. Top options include AMD Radeon RX 5600 XT, AMD Radeon RX 5700 XT, AMD Radeon RX 580 8GB. 114 GPUs have plenty of headroom for comfortable inference.

Which devices can run T5gemma S S Prefixlm?

185 devices with unified memory can run T5gemma S S Prefixlm at BF16 (0.7 GB), including AMD Ryzen 7 8845HS (Hawk Point) Laptop, AMD Ryzen AI 7 350 (Krackan Point) Laptop, AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10. Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.