Gemma · 25.8B parameters · 3.8B active
Gemma 4 26B-A4B: GPU memory requirements and which GPU to rent
The mixture-of-experts Gemma 4: about 26 billion parameters with 3.8 billion active per token and 30 layers, 25 of them on a sliding window — small-model speed, with every expert resident.
- Weights at FP826 GB52 GB at 16-bit · 14 GB at 4-bit
- FP8, 8k context, 1 sequence29 GBweights + KV cache + runtime
- Suggested here1× RTX 5090$279 a month · dedicated
Inference
How much GPU memory Gemma 4 26B-A4B needs to serve
Total memory for one sequence at each context length, and for 16 concurrent sequences at 8k — weights, KV cache at 16-bit and runtime overhead included.
| Weights precision | Weights | 4k context | 32k context | 128k context | 8k × 16 sequences | Suggested here, 8k |
|---|---|---|---|---|---|---|
| 16-bit (BF16 / FP16) | 52 GB | 55 GB | 56 GB | 58 GB | 61 GB | 1× A100 · $649/mo |
| 8-bit (FP8 / INT8) | 26 GB | 29 GB | 29 GB | 31 GB | 34 GB | 1× RTX 5090 · $279/mo |
| 4-bit (GPTQ, AWQ, Q4) | 14 GB | 17 GB | 17 GB | 19 GB | 22 GB | 1× L4 · $169/mo |
GB are 109 bytes. Context lengths are shown for scale where the model supports them; an FP8 KV cache halves the cache part of each figure. As a mixture of experts, every expert stays resident, so memory follows the total parameter count, not the active one. Twenty-five of its thirty layers use a 1,024-token sliding window.
Hardware
GPU servers that run Gemma 4 26B-A4B
For FP8 inference with an 8k context and one sequence: the configurations we would suggest, in order — one large card ahead of several small ones.
Fine-tuning Gemma 4 26B-A4B
Memory for a 2k-token sequence and a micro-batch of 4, with gradient checkpointing. Full fine-tuning keeps gradients and Adam states for every parameter; LoRA and QLoRA train small adapters on a frozen base.
Architecture at a glance
Twenty-five of its thirty layers use a 1,024-token sliding window. Figures from the model's published configuration; method on the calculator page.
Gemma 4 26B-A4B hardware: questions answered
How much VRAM does Gemma 4 26B-A4B need?
For the weights alone: about 52 GB at 16-bit, 26 GB at FP8 and 14 GB at 4-bit. Serving it at FP8 with an 8k-token context and one sequence takes about 29 GB in total once the KV cache and the runtime are counted; long contexts and many concurrent users add the KV cache on top — about 0.23 GB per 1,000 tokens at 16-bit.
Can Gemma 4 26B-A4B run on a single GPU?
Yes, at FP8 with a moderate context: one NVIDIA RTX 5090 (32 GB GDDR7) holds it, for $279 a month here. Longer contexts or more concurrent users may need a larger card or a second GPU.
What is the cheapest GPU setup to run Gemma 4 26B-A4B?
For FP8 inference with an 8k context and one sequence, the configuration we would suggest is 1× RTX 5090 · $279/mo; at 4-bit, 1× L4 · $169/mo. Suggestions favour one large card over several small ones: splitting a model across PCIe cards costs speed.
How much GPU memory do I need to fine-tune Gemma 4 26B-A4B?
With a 2k-token sequence and a micro-batch of 4: about 20 GB for QLoRA on a 4-bit base, 59 GB for LoRA on a 16-bit base, and 419 GB for a full fine-tune with Adam in mixed precision. Gradient checkpointing is assumed; sharding the optimizer across GPUs spreads the full fine-tune over several cards.
Run Gemma 4 26B-A4B on 1× RTX 5090 for $279 a month
Dedicated, single-tenant and live in under 10 minutes — with vLLM, the Hugging Face stack or Ollama preinstalled if you want. Paid in BTC, ETH or USDT, with no KYC.