Which needs less VRAM, Qwen3-30B-A3B or Gemma 4 26B A4B?

Gemma 4 26B A4B: about 19.0 GB at Q4_K_M, against 22.3 GB for Qwen3-30B-A3B. At Q8_0 it's 32.5 GB vs 38.3 GB.

Which is faster locally, Qwen3-30B-A3B or Gemma 4 26B A4B?

Decode speed is limited by how many bytes are read per token. Qwen3-30B-A3B reads 3.0B parameters per token and Gemma 4 26B A4B reads 3.8B, so Qwen3-30B-A3B should be faster on the same machine. Our estimate on the GeForce RTX 5090 at Q4 is up to 837 tok/s vs up to 661 tok/s. These are estimates, not benchmarks.

What is the cheapest GPU for Qwen3-30B-A3B and Gemma 4 26B A4B?

At Q4: Intel Arc Pro B70 32GB ($949) for Qwen3-30B-A3B; Intel Arc Pro B70 32GB ($949) for Gemma 4 26B A4B. Prices are dated list-price snapshots.

Can a Mac run Qwen3-30B-A3B and Gemma 4 26B A4B?

The cheapest Mac or mini PC we track for Qwen3-30B-A3B is the Mac mini M5 Pro 64GB ($3,199). For Gemma 4 26B A4B it's the Mac mini M6 32GB ($1,799).

Which is better, Qwen3-30B-A3B or Gemma 4 26B A4B?

We don't score answer quality. This comparison covers memory, context cache, hardware and speed. Pick the one that fits your hardware, then test both on your own prompts.

Qwen3-30B-A3B vs Gemma 4 26B A4B: VRAM and speed for running locally

Side by side: how much memory each model needs at Q4, Q8 and FP16, how the KV cache grows with context, the cheapest GPU and Mac that run each, and estimated speed on 11 common machines.

Gemma 4 26B A4B is lighter: about 19.0 GB at Q4 against 22.3 GB for Qwen3-30B-A3B, so it runs on 25 tracked devices vs 16. Qwen3-30B-A3B should be faster on the same hardware, since it reads 3.0B parameters per token against 3.8B. Which one gives better answers depends on your task, so check benchmarks for that; this page covers what it takes to run them.

Qwen3-30B-A3B vs Gemma 4 26B A4B: specs and memory
	Qwen3-30B-A3B	Gemma 4 26B A4B
Maker	Qwen	google
License	apache-2.0	apache-2.0
Type	Mixture of experts	Mixture of experts
Total parameters	30.5B	25.8B
Active per token	3.0B	3.8B
Memory at Q4_K_M	22.3 GB	19.0 GB
Memory at Q8_0	38.3 GB	32.5 GB
Memory at FP16	71.2 GB	60.4 GB
KV cache growth	98 MB per 1k tokens, FP16	not modeled
Memory at Q4, 32k context	25.1 GB 3.2 GB cache	cache not modeled
Cheapest GPU at Q4	Intel Arc Pro B70 32GB $949	Intel Arc Pro B70 32GB $949
Cheapest Mac / mini PC at Q4	Mac mini M5 Pro 64GB $3,199	Mac mini M6 32GB $1,799
Cheapest GPU at Q8	NVIDIA RTX 6000 Ada 48GB $6,799	NVIDIA RTX 6000 Ada 48GB $6,799
Tracked devices that run it at Q4	16 of 82	25 of 82

Estimated speed on the same hardware

Decode tokens per second at Q4_K_M from memory bandwidth (how we estimate). MoE figures are upper bounds ("up to"). Estimates, not benchmarks.

Estimated tokens per second, Qwen3-30B-A3B vs Gemma 4 26B A4B
Device	Qwen3-30B-A3B	Gemma 4 26B A4B
GeForce RTX 3060 12GB12 GB usable · 360 GB/s · $329	doesn’t fit	doesn’t fit
GeForce RTX 4060 Ti 16GB16 GB usable · 288 GB/s · $499	doesn’t fit	doesn’t fit
GeForce RTX 508016 GB usable · 960 GB/s · $999	doesn’t fit	doesn’t fit
GeForce RTX 309024 GB usable · 936 GB/s · $1,499	tight fit	up to 345
Radeon RX 7900 XTX24 GB usable · 960 GB/s · $999	tight fit	up to 354
GeForce RTX 509032 GB usable · 1792 GB/s · $1,999	up to 837	up to 661
Mac mini M6 24GB18 GB usable · 170 GB/s · $1,299	doesn’t fit	doesn’t fit
Mac mini M6 32GB24 GB usable · 170 GB/s · $1,799	tight fit	up to 59
Mac mini M5 Pro 64GB48 GB usable · 307 GB/s · $3,199	up to 135	up to 107
Framework Desktop (Ryzen AI Max+ 395) 128GB96 GB usable · 256 GB/s · $3,449	up to 113	up to 89
MacBook Pro M5 Max 128GB96 GB usable · 614 GB/s · $6,699	up to 270	up to 213

Hardware that runs only one of them

Run Gemma 4 26B A4B but not Qwen3-30B-A3B: Radeon RX 7900 XTX, GeForce RTX 3090, GeForce RTX 4090, Mac mini M6 32GB, GeForce RTX 3090 Ti, MacBook Pro M6 32GB.

Current priced devices, cheapest first, at Q4 with ~4k context.

Quick answers

Which needs less VRAM, Qwen3-30B-A3B or Gemma 4 26B A4B?: Gemma 4 26B A4B: about 19.0 GB at Q4_K_M, against 22.3 GB for Qwen3-30B-A3B. At Q8_0 it's 32.5 GB vs 38.3 GB.
Which is faster locally, Qwen3-30B-A3B or Gemma 4 26B A4B?: Decode speed is limited by how many bytes are read per token. Qwen3-30B-A3B reads 3.0B parameters per token and Gemma 4 26B A4B reads 3.8B, so Qwen3-30B-A3B should be faster on the same machine. Our estimate on the GeForce RTX 5090 at Q4 is up to 837 tok/s vs up to 661 tok/s. These are estimates, not benchmarks.
What is the cheapest GPU for Qwen3-30B-A3B and Gemma 4 26B A4B?: At Q4: Intel Arc Pro B70 32GB ($949) for Qwen3-30B-A3B; Intel Arc Pro B70 32GB ($949) for Gemma 4 26B A4B. Prices are dated list-price snapshots.
Can a Mac run Qwen3-30B-A3B and Gemma 4 26B A4B?: The cheapest Mac or mini PC we track for Qwen3-30B-A3B is the Mac mini M5 Pro 64GB ($3,199). For Gemma 4 26B A4B it's the Mac mini M6 32GB ($1,799).
Which is better, Qwen3-30B-A3B or Gemma 4 26B A4B?: We don't score answer quality. This comparison covers memory, context cache, hardware and speed. Pick the one that fits your hardware, then test both on your own prompts.

Related comparisons

gpt-oss-20b vs Qwen3-30B-A3B15.6 vs 22.3 GB

All comparisons · MoE vs dense · Q4 vs Q8 vs FP16 · Context length and VRAM · VRAM calculator

Qwen3-30B-A3B vs Gemma 4 26B A4B: VRAM and speed for running locally

Estimated speed on the same hardware

Hardware that runs only one of them

Quick answers

Related comparisons

New models, and what they need to run