Gemma 4 31B IT vs Qwen3 30B A3B Thinking 2507
Gemma 4 31B IT leads llmrun Score 106.1 to 100.6 (ranges overlap); both fit 24 GB at Q4_K_M; Qwen3 30B A3B Thinking 2507 is 5.4× faster on RTX 4090 (~173 vs ~32 tok/s).
Model A · Google
Gemma 4 31B IT31.3B · 256K context · Apache 2.0 · Mar 2026
Model B · Alibaba
Qwen3 30B A3B Thinking 250730.5B · 3.4B active · 256K context · Apache 2.0 · Jul 2025
Choose Gemma 4 31B IT if you want…
Higher llmrun Score
106.1 vs 100.6, ranges overlap
Stronger at reasoning
109.3 vs 93.6
Stronger at science
104.8 vs 98.8
Ahead on the benchmarks
6 of 7 shared boards
Also does vision
Qwen3 30B A3B Thinking 2507 doesn't list it
Newer release
Mar 2026 vs Jul 2025
Choose Qwen3 30B A3B Thinking 2507 if you want…
5.4× faster on RTX 4090
~173 vs ~32 tok/s
Also does chat
Gemma 4 31B IT doesn't list it
On your hardware
Each model at its default quant (Q4_K_M), a smaller one when it must. Estimates, ±20%.
8 GB
RTX 4060 · 8 GB tier
Gemma 4 31B IT
Doesn't fit
smallest quant needs 11.5 GB
Qwen3 30B A3B Thinking 2507
Doesn't fit
smallest quant needs 9.5 GB
12 GB
RTX 3060 12GB · 12 GB tier
Gemma 4 31B IT
IQ2_XXS · ~23 tok/s
11.5 GB · up to 16K
Qwen3 30B A3B Thinking 2507
IQ2_M · ~130 tok/s
11.4 GB · up to 16K
16 GB
RTX 5060 Ti 16GB · 16 GB tier
Gemma 4 31B IT
IQ3_XS · ~20 tok/s
15.8 GB · up to 16K
Qwen3 30B A3B Thinking 2507
Q3_K_M · ~119 tok/s
16 GB · up to 16K
24 GBreference
RTX 4090 · 24 GB tier
Gemma 4 31B IT
Q4_K_M · ~32 tok/s
21.7 GB · up to 32K
Qwen3 30B A3B Thinking 2507
Q4_K_M · ~173 tok/s
19.4 GB · up to 64Kfaster
32 GB
RTX 5090 · 32 GB tier
Gemma 4 31B IT
Q4_K_M · ~57 tok/s
21.7 GB · up to 64K
Qwen3 30B A3B Thinking 2507
Q4_K_M · ~227 tok/s
19.4 GB · up to 256Kfaster
48 GB
RTX A6000 · 48 GB tier
Gemma 4 31B IT
Q4_K_M · ~25 tok/s
21.7 GB · up to 128K
Qwen3 30B A3B Thinking 2507
Q4_K_M · ~148 tok/s
19.4 GB · up to 256Kfaster
Mac 32 GB
Mac Mini M4 · Macs
Gemma 4 31B IT
Q4_K_M · ~4 tok/s
21.7 GB · up to 64K
Qwen3 30B A3B Thinking 2507
Q4_K_M · ~29 tok/s
19.4 GB · up to 128Kfaster
Mac 64 GB
Mac Studio M4 Max · Macs
Gemma 4 31B IT
Q4_K_M · ~19 tok/s
21.7 GB · up to 256K
Qwen3 30B A3B Thinking 2507
Q4_K_M · ~67 tok/s
19.4 GB · up to 256Kfaster
Mac 128 GB
Mac Studio M4 Max · Macs
Gemma 4 31B IT
Q4_K_M · ~19 tok/s
21.7 GB · up to 256K
Qwen3 30B A3B Thinking 2507
Q4_K_M · ~67 tok/s
19.4 GB · up to 256Kfaster
Reference GPU: RTX 4090. Tinted cell = the better pick on that machine (fits, more headroom, or at least 10% faster).
Overall
Gemma 4 31B IT
Qwen3 30B A3B Thinking 2507
Coding
Gemma 4 31B IT
Qwen3 30B A3B Thinking 2507
no data
Science
Gemma 4 31B IT
Qwen3 30B A3B Thinking 2507
Reasoning
Gemma 4 31B IT
Qwen3 30B A3B Thinking 2507
Gemma 4 31B IT · #35 of 92 open models · #4 of 34 that fit 24 GB
Qwen3 30B A3B Thinking 2507 · #45 of 92 open models · #9 of 34 that fit 24 GB
Benchmarks
7 shared boards: Gemma 4 31B IT leads 6, Qwen3 30B A3B Thinking 2507 leads 1 · Benchmarks imported Oct 2026
Gemma 4 31B IT
75.8%#27/97 open
IndependentEpoch AI · Apr 2026
Qwen3 30B A3B Thinking 2507
70.1%#32/97 open
IndependentEpoch AI · Jul 2025
Gemma 4 31B IT
73.3%#25/85 open
IndependentEpoch AI · Apr 2026
Qwen3 30B A3B Thinking 2507
70.3%#27/85 open
IndependentEpoch AI · Jul 2025
Gemma 4 31B IT
43.4%#23/65 open
Lab-reportedEpoch AI · Apr 2026
Qwen3 30B A3B Thinking 2507
33.3%#48/65 open
Lab-reportedEpoch AI · Jul 2025
Gemma 4 31B IT
5.0%#29/67 open
IndependentEpoch AI · Apr 2026
Qwen3 30B A3B Thinking 2507
8.0%#26/67 open
IndependentEpoch AI · Jul 2025
Gemma 4 31B IT
1.4%#29/65 open
Lab-reportedEpoch AI · Apr 2026
Qwen3 30B A3B Thinking 2507
0.3%#41/65 open
Lab-reportedEpoch AI · Jul 2025
Gemma 4 31B IT
82.7%#19/68 open
Lab-reportedEpoch AI · Apr 2026
Qwen3 30B A3B Thinking 2507
69.3%#33/68 open
Lab-reportedEpoch AI · Jul 2025
Gemma 4 31B IT
39.3%#9/54 open
Lab-reportedEpoch AI · Apr 2026
Qwen3 30B A3B Thinking 2507
19.5%#34/54 open
Lab-reportedEpoch AI · Jul 2025
7 more boards (other shared and single-model boards)
Only Gemma 4 31B IT (6)
- LMArena Text1452.4
- LMArena WebDev1365.9
- SimpleQA10.4%
- WeirdML52.3%
- ALE-Bench925.5
- Surface Evolver Bench30.6%
Only Qwen3 30B A3B Thinking 2507 (1)
- MMLU-Pro80.9%
VRAM by quantization
Memory to load the model and 2K context; below it, the total at 32K and 128K context. Lighter side in bold.
Q2_K
Gemma 4 31B IT
14.9 GB12.6 GB file
32K 18.4 GB · 128K 31.0 GB
Qwen3 30B A3B Thinking 2507
13.4 GB11.3 GB file
32K 14.9 GB · 128K 19.7 GB
Q3_K_S
Gemma 4 31B IT
15.3 GB13.2 GB file
32K 18.7 GB · 128K 31.4 GB
Qwen3 30B A3B Thinking 2507
13.8 GB13.3 GB file
32K 15.3 GB · 128K 20.1 GB
Q3_K_M
Gemma 4 31B IT
16.9 GB14.7 GB file
32K 20.3 GB · 128K 33.0 GB
Qwen3 30B A3B Thinking 2507
15.3 GB14.7 GB file
32K 16.8 GB · 128K 21.6 GB
Q4_0
Gemma 4 31B IT
17.3 GB17.3 GB file
32K 20.7 GB · 128K 33.4 GB
Qwen3 30B A3B Thinking 2507
15.7 GB17.4 GB file
32K 17.2 GB · 128K 22.0 GB
Q4_K_M
Gemma 4 31B IT
20.4 GB18.3 GB file
32K 23.8 GB · 128K 36.5 GB
Qwen3 30B A3B Thinking 2507
18.7 GB18.6 GB file
32K 20.2 GB · 128K 25.1 GB
Q5_K_M
Gemma 4 31B IT
23.9 GB21.7 GB file
32K 27.3 GB · 128K 40.0 GB
Qwen3 30B A3B Thinking 2507
22.2 GB21.7 GB file
32K 23.7 GB · 128K 28.5 GB
Q6_K
Gemma 4 31B IT
27.4 GB25.2 GB file
32K 30.9 GB · 128K 43.5 GB
Qwen3 30B A3B Thinking 2507
25.6 GB25.1 GB file
32K 27.1 GB · 128K 31.9 GB
Q8_0
Gemma 4 31B IT
32.9 GB~31.3 GB file
32K 36.3 GB · 128K 49.0 GB
Qwen3 30B A3B Thinking 2507
30.9 GB32.5 GB file
32K 32.4 GB · 128K 37.3 GB
19 more quants
IQ2_XXS
Gemma 4 31B IT
10.2 GB8.5 GB file
32K 13.7 GB · 128K 26.3 GB
Qwen3 30B A3B Thinking 2507
8.8 GB10.3 GB file
32K 10.3 GB · 128K 15.1 GB
IQ2_XS
Gemma 4 31B IT
11.0 GB11.5 GB file
32K 14.4 GB · 128K 27.1 GB
Qwen3 30B A3B Thinking 2507
9.6 GB8.7 GB file
32K 11.1 GB · 128K 15.9 GB
IQ2_S
Gemma 4 31B IT
11.4 GB12.1 GB file
32K 14.8 GB · 128K 27.5 GB
Qwen3 30B A3B Thinking 2507
9.9 GB8.7 GB file
32K 11.5 GB · 128K 16.3 GB
IQ2_M
Gemma 4 31B IT
12.2 GB10.8 GB file
32K 15.6 GB · 128K 28.3 GB
Qwen3 30B A3B Thinking 2507
10.7 GB10.8 GB file
32K 12.2 GB · 128K 17.1 GB
IQ3_XXS
Gemma 4 31B IT
13.7 GB11.8 GB file
32K 17.2 GB · 128K 29.9 GB
Qwen3 30B A3B Thinking 2507
12.2 GB12.9 GB file
32K 13.7 GB · 128K 18.6 GB
Q2_K_S
Gemma 4 31B IT
14.1 GB11.0 GB file
32K 17.6 GB · 128K 30.3 GB
Qwen3 30B A3B Thinking 2507
12.6 GB10.5 GB file
32K 14.1 GB · 128K 19.0 GB
IQ3_XS
Gemma 4 31B IT
14.5 GB13.8 GB file
32K 18.0 GB · 128K 30.6 GB
Qwen3 30B A3B Thinking 2507
13.0 GB12.7 GB file
32K 14.5 GB · 128K 19.3 GB
IQ3_S
Gemma 4 31B IT
14.9 GB13.8 GB file
32K 18.4 GB · 128K 31.0 GB
Qwen3 30B A3B Thinking 2507
13.4 GB13.3 GB file
32K 14.9 GB · 128K 19.7 GB
IQ3_M
Gemma 4 31B IT
15.7 GB15.1 GB file
32K 19.1 GB · 128K 31.8 GB
Qwen3 30B A3B Thinking 2507
14.1 GB14.1 GB file
32K 15.7 GB · 128K 20.5 GB
Q3_K_L
Gemma 4 31B IT
17.7 GB16.8 GB file
32K 21.1 GB · 128K 33.8 GB
Qwen3 30B A3B Thinking 2507
16.1 GB14.6 GB file
32K 17.6 GB · 128K 22.4 GB
IQ4_XS
Gemma 4 31B IT
18.4 GB16.4 GB file
32K 21.9 GB · 128K 34.6 GB
Qwen3 30B A3B Thinking 2507
16.8 GB16.4 GB file
32K 18.3 GB · 128K 23.2 GB
IQ4_NL
Gemma 4 31B IT
19.2 GB17.3 GB file
32K 22.7 GB · 128K 35.3 GB
Qwen3 30B A3B Thinking 2507
17.6 GB17.3 GB file
32K 19.1 GB · 128K 23.9 GB
Q4_1
Gemma 4 31B IT
19.2 GB19.1 GB file
32K 22.7 GB · 128K 35.3 GB
Qwen3 30B A3B Thinking 2507
17.6 GB19.2 GB file
32K 19.1 GB · 128K 23.9 GB
Q4_K_S
Gemma 4 31B IT
19.2 GB17.4 GB file
32K 22.7 GB · 128K 35.3 GB
Qwen3 30B A3B Thinking 2507
17.6 GB17.5 GB file
32K 19.1 GB · 128K 23.9 GB
Q4_K_L
Gemma 4 31B IT
20.8 GB19.9 GB file
32K 24.2 GB · 128K 36.9 GB
Qwen3 30B A3B Thinking 2507
19.1 GB18.9 GB file
32K 20.6 GB · 128K 25.4 GB
Q5_0
Gemma 4 31B IT
21.2 GB21.3 GB file
32K 24.6 GB · 128K 37.3 GB
Qwen3 30B A3B Thinking 2507
19.5 GB21.1 GB file
32K 21.0 GB · 128K 25.8 GB
Q5_K_S
Gemma 4 31B IT
23.1 GB21.1 GB file
32K 26.6 GB · 128K 39.2 GB
Qwen3 30B A3B Thinking 2507
21.4 GB21.1 GB file
32K 22.9 GB · 128K 27.7 GB
Q5_K_L
Gemma 4 31B IT
24.3 GB23.0 GB file
32K 27.7 GB · 128K 40.4 GB
Qwen3 30B A3B Thinking 2507
22.5 GB21.9 GB file
32K 24.1 GB · 128K 28.9 GB
BF16
Gemma 4 31B IT
64.2 GB~62.5 GB file
32K 67.6 GB · 128K 80.3 GB
Qwen3 30B A3B Thinking 2507
61.5 GB61.1 GB file
32K 63.0 GB · 128K 67.8 GB
Gemma 4 31B IT
Ollama · Q4_K_M
ollama run hf.co/unsloth/gemma-4-31B-it-GGUF:Q4_K_MQwen3 30B A3B Thinking 2507
Ollama · Q4_K_M
ollama run hf.co/unsloth/Qwen3-30B-A3B-Thinking-2507-GGUF:Q4_K_M| Spec | Gemma 4 31B IT | Qwen3 30B A3B Thinking 2507 |
|---|---|---|
| Publisher | Alibaba | |
| Parameters | 31.3B | 30.5B |
| Active parameters | — | 3.4B per token |
| Architecture | Gemma4ForConditionalGeneration | Qwen3MoeForCausalLM |
| Context | 256K tokens | 256K tokens |
| Modalities | Vision, Function-calling | Chat, Function-calling |
| License | Apache 2.0 | Apache 2.0 |
| Released | Mar 2026 | Jul 2025 |
Frequently Asked Questions
Which is better, Gemma 4 31B IT or Qwen3 30B A3B Thinking 2507?
Gemma 4 31B IT leads llmrun Score 106.1 to 100.6 (ranges overlap); both fit 24 GB at Q4_K_M; Qwen3 30B A3B Thinking 2507 is 5.4× faster on RTX 4090 (~173 vs ~32 tok/s). Choose Gemma 4 31B IT for higher llmrun Score and stronger at reasoning; choose Qwen3 30B A3B Thinking 2507 for 5.4× faster on RTX 4090 and also does chat.
Which needs less VRAM, Gemma 4 31B IT or Qwen3 30B A3B Thinking 2507?
At Q4_K_M, Gemma 4 31B IT requires 20.4 GB while Qwen3 30B A3B Thinking 2507 needs 18.7 GB, making Qwen3 30B A3B Thinking 2507 the easier model to run on your local setup.
Which is faster on RTX 4090?
Estimated decode speed on RTX 4090 is about 32 tok/s for Gemma 4 31B IT (Q4_K_M) and 173 tok/s for Qwen3 30B A3B Thinking 2507 (Q4_K_M), so Qwen3 30B A3B Thinking 2507 is faster. Estimates, ±20%.
Can I run Gemma 4 31B IT or Qwen3 30B A3B Thinking 2507 on a 16 GB GPU?
Gemma 4 31B IT fits at IQ3_XS; Qwen3 30B A3B Thinking 2507 fits at Q3_K_M.
Which has a longer context window, Gemma 4 31B IT or Qwen3 30B A3B Thinking 2507?
Gemma 4 31B IT supports 262,144 tokens and Qwen3 30B A3B Thinking 2507 supports 262,144 tokens.