Can I run gemma-4-31B on an RTX 5090?

Probably, at 4-bit

The published weights are too big, but a 4-bit version (about 17.6 GB) would fit, with room for ≈ 62K tokens of context. The site hasn't analyzed a GGUF of gemma-4-31B yet.

Full analysis of gemma-4-31B Everything that runs on RTX 5090

Files and context on RTX 5090

Each file of gemma-4-31B (and of GGUF versions the site has analyzed), and how much context fits next to it, with the context cache in 16-bit or 8-bit.

File	Size	On RTX 5090	Context (16-bit)	Context (8-bit)
As published	62.5 GB	Doesn't fit	—	—
4-bit (estimate)	17.6 GB	Fits	62K	129K

gemma-4-31B on other hardware

NVIDIA H100Yes, just
RTX 5090Probably, at 4-bit
RTX 4090Probably, at 4-bit
RTX 3090Probably, at 4-bit
RTX 5060 Ti 16 GBNo
RTX 3060 12 GBNo
RTX 4060No
Mac with 32 GBProbably, at 4-bit
Mac with 64 GBProbably, at 4-bit
Mac with 128 GBYes

Popular models that run on RTX 5090

Qwen3-0.6BQwen · 752M
Qwen3-VL-8B-InstructQwen · 8.77B
gpt2openai-community · 137M
tiny-Qwen2ForCausalLM-2.5trl-internal-testing · 2.44M
Qwen3-8BQwen · 8.19B
Qwen3-Coder-30B-A3B-Instruct-GGUFunsloth · 30.5B
Qwen2.5-7B-InstructQwen · 7.62B
Qwen3.5-9BQwen · 9.65B

Assumes 90% of the card's memory is usable and at least 4K tokens of context for one conversation, with the context cache in 16-bit unless noted. Fits leaves at least 20% of that memory for context; Tight leaves less.