Skip to content
Local AI Wizard
Your machine
Theme

Gemma 4 26B A4B vs Qwen3 30B A3B for local chat

Two mixture-of-experts chat models for a 24 to 32 GB machine

Facts from the reviewed catalog

Facts from the reviewed catalog
SpecificationGemma 4 26B A4BQwen3 30B A3B Instruct 2507
TaskText & chatText & chat
Parameters25.2B · 3.8B active30.5B · 3.3B active
LicenseApache 2.0Apache 2.0
Commercial useAllowedlicense textAllowedlicense text
Quality · LMArena 2026-10-22 · text1,437 ± 71,430–1,445 · 6,092 votes · rank 100 · source1,382 ± 51,377–1,387 · 23,306 votes · rank 190 · source
MemoryQ4 planning estimate (~19 GB of weights before the machine's reserve)Q4 planning estimate (~23 GB of weights before the machine's reserve)
Checkpointofficial cardofficial card

In LMArena 2026-10-22 · text, Gemma 4 26B A4B rates 1,437 ± 7 against 1,382 ± 5 for Qwen3 30B A3B Instruct 2507, 55 points ahead with intervals that do not overlap. Blind pairwise votes by people on the hosted model, CC BY 4.0. A rating is relative; overlapping intervals mean the order is not settled. Scale: Arena rating (style control, overall) ± half the confidence interval.

Memory on three machines

Each machine's own reserve is added (graphics cards 2 GB; unified and system memory 25%, 2 to 8 GB). Planning estimates, not measured runs.

Gemma 4 26B A4B vs Qwen3 30B A3B for local chat: memory on three machines
MachineGemma 4 26B A4BQwen3 30B A3B Instruct 2507
RTX 4090 · 24 GB VRAM~21 GBFast memory holds it~25 GBIn memory, outside the fast part · offload to PC RAM (32 GB assumed)
RTX 5090 · 32 GB VRAM~21 GBFast memory holds it~25 GBFast memory holds it
MacBook Air 13-inch M4 · 32 GB unified~27 GBFast memory holds it~31 GBFast memory holds it

Gemma 4 26B A4B: Q4 planning estimate (~19 GB of weights before the machine's reserve). Qwen3 30B A3B Instruct 2507: Q4 planning estimate (~23 GB of weights before the machine's reserve). Formulas

Everything that runs on each machine: RTX 4090 · 24 GB VRAM · RTX 5090 · 32 GB VRAM · MacBook Air 13-inch M4 · 32 GB unified

Which to choose

Both activate about 4 billion parameters per token or fewer (3.8 billion for Gemma 4, 3.3 billion for Qwen3), so both generate quickly for their size, and both are Apache 2.0 with a 262,144-token context. Gemma 4 26B A4B is a little smaller in total (25.2 billion against 30.5), so its Q4 weights take about 19 GB against 23 GB, and it reads images; Qwen3 30B A3B Instruct 2507 is text only. In the same LMArena text release Gemma 4 26B A4B rates higher.

Files and run commands

The default file of the most downloaded collected source, or the commands each card documents. Not run by this site.

Gemma 4 26B A4B

unsloth/gemma-4-26B-A4B-it-GGUF · gemma-4-26B-A4B-it-UD-Q4_K_M.gguf · UD-Q4_K_M · 16.9 GB · unsloth

ollama run hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_M
llama-server -hf unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_M
lms get https://huggingface.co/unsloth/gemma-4-26B-A4B-it-GGUF
  • mlx-lm: mlx-lm runs MLX repositories; pick the MLX version on a Mac

Qwen3 30B A3B Instruct 2507

unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF · Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf · Q4_K_M · 18.6 GB · unsloth

ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_M
llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_M
lms get https://huggingface.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF
  • mlx-lm: mlx-lm runs MLX repositories; pick the MLX version on a Mac

Numbers each publisher reports

Each card's own results with its own settings. Publishers choose different tests and settings, so these values are never compared across publishers here.

Gemma 4 26B A4B · reported by Google

15 results · Google's model card · read 2026-10-25
MMLU Pro
82.6%
AIME 2026 no tools
88.3%
LiveCodeBench v6
77.1%
Codeforces ELO
1,718
GPQA Diamond
82.3%
Tau2 (average over 3)
68.2%
HLE no tools
8.7%
HLE with search
17.2%
Every reported result15
BigBench Extra Hard
64.8%
MMMLU
86.3%
MMMU Pro
73.8%Vision
OmniDocBench 1.5 (average edit distance, lower is better)
0.149Vision
MATH-Vision
82.4%Vision
MedXPertQA MM
58.1%Vision
MRCR v2 8 needle 128k (average)
44.1%Long Context

Values as printed on the card, with the publisher's own settings (card section: Benchmark Results); column “Gemma 4 26B A4B”. Not comparable across publishers and not used in any ranking. Model card

Qwen3 30B A3B Instruct 2507 · reported by Alibaba

25 results · Alibaba's model card · read 2026-10-25
MMLU-Pro
78.4Knowledge
MMLU-Redux
89.3Knowledge
GPQA
70.4Knowledge
SuperGPQA
53.4Knowledge
AIME25
61.3Reasoning
HMMT25
43.0Reasoning
ZebraLogic
90.0Reasoning
LiveBench 20241125
69.0Reasoning
Every reported result25
LiveCodeBench v6 (25.02-25.05)
43.2Coding
MultiPL-E
83.8Coding
Aider-Polyglot
35.6Coding
IFEval
84.7Alignment
Arena-Hard v2
69.0Alignment
Creative Writing v3
86.0Alignment
WritingBench
85.5Alignment
BFCL-v3
65.1Agent
TAU1-Retail
59.1Agent
TAU1-Airline
40.0Agent
TAU2-Retail
57.0Agent
TAU2-Airline
38.0Agent
TAU2-Telecom
12.3Agent
MultiIF
67.9Multilingualism
MMLU-ProX
72.0Multilingualism
INCLUDE
71.9Multilingualism
PolyMATH
43.1Multilingualism

Values as printed on the card, with the publisher's own settings (card section: Performance); column “Qwen3-30B-A3B-Instruct-2507”. Not comparable across publishers and not used in any ranking. Model card

Limits and dates

The published quality benchmark does not test a local quantized build. Memory estimates are not speed measurements or guaranteed hardware fits. Repository size, when available, is not VRAM. Formulas

Page updated · Catalog reviewed · LMArena 2026-10-22 · text