Qwen3 30B A3B Instruct 2507 vs Qwen3.8 27B
Qwen3.8 27B leads llmrun Score 121.5 to 96.5; both fit 24 GB at Q4_K_M; Qwen3 30B A3B Instruct 2507 is 4.6× faster on RTX 4090 (~173 vs ~38 tok/s).
Model A · Alibaba
Qwen3 30B A3B Instruct 250730.5B · 3.4B active · 256K context · Apache 2.0 · Jul 2025
Model B · Alibaba
Qwen3.8 27B27.8B · 256K context · Apache 2.0 · Aug 2026
Choose Qwen3 30B A3B Instruct 2507 if you want…
4.6× faster on RTX 4090
~173 vs ~38 tok/s
Also does chat
Qwen3.8 27B doesn't list it
Choose Qwen3.8 27B if you want…
Higher llmrun Score
121.5 vs 96.5
Stronger at reasoning
124.1 vs 91.9
Ahead on the benchmarks
3 of 3 shared boards
Fits a 12 GB GPU
IQ3_XXS · 12 GB; Qwen3 30B A3B Instruct 2507 needs 9.5 GB at its smallest quant
Also does vision
Qwen3 30B A3B Instruct 2507 doesn't list it
Newer release
Aug 2026 vs Jul 2025
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
Qwen3 30B A3B Instruct 2507
Doesn't fit
smallest quant needs 9.5 GB
Qwen3.8 27B
Doesn't fit
smallest quant needs 8.8 GB
12 GB
RTX 3060 12GB · 12 GB tier
Qwen3 30B A3B Instruct 2507
IQ2_M · ~130 tok/s
11.4 GB · up to 16K
Qwen3.8 27B
IQ3_XXS · ~20 tok/s
12 GB · up to 16Kbetter quant
16 GB
RTX 5060 Ti 16GB · 16 GB tier
Qwen3 30B A3B Instruct 2507
Q3_K_M · ~119 tok/s
16 GB · up to 16K
Qwen3.8 27B
Q3_K_L · ~19 tok/s
15.4 GB · up to 16K
24 GBreference
RTX 4090 · 24 GB tier
Qwen3 30B A3B Instruct 2507
Q4_K_M · ~173 tok/s
19.4 GB · up to 64Kfaster
Qwen3.8 27B
Q4_K_M · ~38 tok/s
17.9 GB · up to 64K
32 GB
RTX 5090 · 32 GB tier
Qwen3 30B A3B Instruct 2507
Q4_K_M · ~227 tok/s
19.4 GB · up to 256Kfaster
Qwen3.8 27B
Q4_K_M · ~67 tok/s
17.9 GB · up to 256K
48 GB
RTX A6000 · 48 GB tier
Qwen3 30B A3B Instruct 2507
Q4_K_M · ~148 tok/s
19.4 GB · up to 256Kfaster
Qwen3.8 27B
Q4_K_M · ~29 tok/s
17.9 GB · up to 256K
Mac 32 GB
Mac Mini M4 · Macs
Qwen3 30B A3B Instruct 2507
Q4_K_M · ~29 tok/s
19.4 GB · up to 128Kfaster
Qwen3.8 27B
Q4_K_M · ~5 tok/s
17.9 GB · up to 128K
Mac 64 GB
Mac Studio M4 Max · Macs
Qwen3 30B A3B Instruct 2507
Q4_K_M · ~67 tok/s
19.4 GB · up to 256Kfaster
Qwen3.8 27B
Q4_K_M · ~22 tok/s
17.9 GB · up to 256K
Mac 128 GB
Mac Studio M4 Max · Macs
Qwen3 30B A3B Instruct 2507
Q4_K_M · ~67 tok/s
19.4 GB · up to 256Kfaster
Qwen3.8 27B
Q4_K_M · ~22 tok/s
17.9 GB · up to 256K
Reference GPU: RTX 4090. Tinted cell = the better pick on that machine (fits, more headroom, or at least 10% faster).
Overall
Qwen3 30B A3B Instruct 2507
Qwen3.8 27B
Coding
Qwen3 30B A3B Instruct 2507
no data
Qwen3.8 27B
Math
Qwen3 30B A3B Instruct 2507
no data
Qwen3.8 27B
Science
Qwen3 30B A3B Instruct 2507
no data
Qwen3.8 27B
Reasoning
Qwen3 30B A3B Instruct 2507
Qwen3.8 27B
Qwen3 30B A3B Instruct 2507 · #56 of 92 open models · #14 of 34 that fit 24 GB
Qwen3.8 27B · #10 of 92 open models · #1 of 34 that fit 24 GB
Benchmarks
3 shared boards: Qwen3 30B A3B Instruct 2507 leads 0, Qwen3.8 27B leads 3 · Benchmarks imported Oct 2026
Qwen3 30B A3B Instruct 2507
1382.4#59/186 open
ArenaLMArena · Oct 2026
Qwen3.8 27B
1438.1#22/186 open
ArenaLMArena · Oct 2026
Qwen3 30B A3B Instruct 2507
67.2%#36/68 open
Lab-reportedEpoch AI · Jul 2025
Qwen3.8 27B
88.0%#9/68 open
Lab-reportedEpoch AI · Aug 2026
Qwen3 30B A3B Instruct 2507
22.4%#31/54 open
Lab-reportedEpoch AI · Jul 2025
Qwen3.8 27B
41.4%#7/54 open
Lab-reportedEpoch AI · Aug 2026
14 more boards (other shared and single-model boards)
Only Qwen3 30B A3B Instruct 2507 (3)
- GPQA Diamond55.6%
- AIME 2024/202562.2%
- Chess Puzzles2.0%
Only Qwen3.8 27B (11)
- LiveBench Coding75.7
- LiveBench Reasoning80.0
- LMArena WebDev1592.8
- LiveBench Math86.2
- ARC-AGI-242.4%
- SciCode46.6%
- ARC-AGI87.5%
- CritPt5.4%
- APEX-Agents47.5%
- Surface Evolver Bench45.0%
- ProofBench16.0%
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
Qwen3 30B A3B Instruct 2507
13.4 GB11.3 GB file
32K 14.9 GB · 128K 19.7 GB
Qwen3.8 27B
12.6 GB10.8 GB file
32K 13.9 GB · 128K 19.3 GB
Q3_K_S
Qwen3 30B A3B Instruct 2507
13.8 GB13.3 GB file
32K 15.3 GB · 128K 20.1 GB
Qwen3.8 27B
12.9 GB12.7 GB file
32K 14.2 GB · 128K 19.6 GB
Q3_K_M
Qwen3 30B A3B Instruct 2507
15.3 GB14.7 GB file
32K 16.8 GB · 128K 21.6 GB
Qwen3.8 27B
14.3 GB13.4 GB file
32K 15.6 GB · 128K 21.0 GB
Q4_0
Qwen3 30B A3B Instruct 2507
15.7 GB17.4 GB file
32K 17.2 GB · 128K 22.0 GB
Qwen3.8 27B
14.6 GB~13.9 GB file
32K 16.0 GB · 128K 21.4 GB
Q4_K_M
Qwen3 30B A3B Instruct 2507
18.7 GB18.6 GB file
32K 20.2 GB · 128K 25.1 GB
Qwen3.8 27B
17.4 GB16.5 GB file
32K 18.8 GB · 128K 24.1 GB
Q5_K_M
Qwen3 30B A3B Instruct 2507
22.2 GB21.7 GB file
32K 23.7 GB · 128K 28.5 GB
Qwen3.8 27B
20.5 GB19.8 GB file
32K 21.9 GB · 128K 27.3 GB
Q6_K
Qwen3 30B A3B Instruct 2507
25.6 GB25.1 GB file
32K 27.1 GB · 128K 31.9 GB
Qwen3.8 27B
23.7 GB22.0 GB file
32K 25.0 GB · 128K 30.4 GB
Q8_0
Qwen3 30B A3B Instruct 2507
30.9 GB32.5 GB file
32K 32.4 GB · 128K 37.3 GB
Qwen3.8 27B
28.5 GB29.0 GB file
32K 29.9 GB · 128K 35.2 GB
19 more quants
IQ2_XXS
Qwen3 30B A3B Instruct 2507
8.8 GB10.3 GB file
32K 10.3 GB · 128K 15.1 GB
Qwen3.8 27B
8.4 GB7.3 GB file
32K 9.7 GB · 128K 15.1 GB
IQ2_XS
Qwen3 30B A3B Instruct 2507
9.6 GB8.7 GB file
32K 11.1 GB · 128K 15.9 GB
Qwen3.8 27B
9.1 GB9.1 GB file
32K 10.4 GB · 128K 15.8 GB
IQ2_S
Qwen3 30B A3B Instruct 2507
9.9 GB8.7 GB file
32K 11.5 GB · 128K 16.3 GB
Qwen3.8 27B
9.4 GB8.4 GB file
32K 10.8 GB · 128K 16.1 GB
IQ2_M
Qwen3 30B A3B Instruct 2507
10.7 GB10.8 GB file
32K 12.2 GB · 128K 17.1 GB
Qwen3.8 27B
10.1 GB10.5 GB file
32K 11.5 GB · 128K 16.8 GB
IQ3_XXS
Qwen3 30B A3B Instruct 2507
12.2 GB12.9 GB file
32K 13.7 GB · 128K 18.6 GB
Qwen3.8 27B
11.5 GB10.9 GB file
32K 12.9 GB · 128K 18.2 GB
Q2_K_S
Qwen3 30B A3B Instruct 2507
12.6 GB10.5 GB file
32K 14.1 GB · 128K 19.0 GB
Qwen3.8 27B
not listed
IQ3_XS
Qwen3 30B A3B Instruct 2507
13.0 GB12.7 GB file
32K 14.5 GB · 128K 19.3 GB
Qwen3.8 27B
12.2 GB10.4 GB file
32K 13.6 GB · 128K 18.9 GB
IQ3_S
Qwen3 30B A3B Instruct 2507
13.4 GB10.3 GB file
32K 14.9 GB · 128K 19.7 GB
Qwen3.8 27B
12.6 GB12.0 GB file
32K 13.9 GB · 128K 19.3 GB
IQ3_M
Qwen3 30B A3B Instruct 2507
14.1 GB14.1 GB file
32K 15.7 GB · 128K 20.5 GB
Qwen3.8 27B
13.3 GB14.9 GB file
32K 14.6 GB · 128K 20.0 GB
Q3_K_L
Qwen3 30B A3B Instruct 2507
16.1 GB15.9 GB file
32K 17.6 GB · 128K 22.4 GB
Qwen3.8 27B
15.0 GB14.1 GB file
32K 16.3 GB · 128K 21.7 GB
IQ4_XS
Qwen3 30B A3B Instruct 2507
16.8 GB16.4 GB file
32K 18.3 GB · 128K 23.2 GB
Qwen3.8 27B
15.7 GB14.3 GB file
32K 17.0 GB · 128K 22.4 GB
IQ4_NL
Qwen3 30B A3B Instruct 2507
17.6 GB17.3 GB file
32K 19.1 GB · 128K 23.9 GB
Qwen3.8 27B
16.4 GB17.4 GB file
32K 17.7 GB · 128K 23.1 GB
Q4_1
Qwen3 30B A3B Instruct 2507
17.6 GB19.2 GB file
32K 19.1 GB · 128K 23.9 GB
Qwen3.8 27B
16.4 GB17.5 GB file
32K 17.7 GB · 128K 23.1 GB
Q4_K_S
Qwen3 30B A3B Instruct 2507
17.6 GB17.5 GB file
32K 19.1 GB · 128K 23.9 GB
Qwen3.8 27B
16.4 GB15.4 GB file
32K 17.7 GB · 128K 23.1 GB
Q4_K_L
Qwen3 30B A3B Instruct 2507
19.1 GB18.9 GB file
32K 20.6 GB · 128K 25.4 GB
Qwen3.8 27B
17.8 GB18.8 GB file
32K 19.1 GB · 128K 24.5 GB
Q5_0
Qwen3 30B A3B Instruct 2507
19.5 GB21.1 GB file
32K 21.0 GB · 128K 25.8 GB
Qwen3.8 27B
not listed
Q5_K_S
Qwen3 30B A3B Instruct 2507
21.4 GB21.1 GB file
32K 22.9 GB · 128K 27.7 GB
Qwen3.8 27B
19.9 GB18.7 GB file
32K 21.2 GB · 128K 26.6 GB
Q5_K_L
Qwen3 30B A3B Instruct 2507
22.5 GB21.9 GB file
32K 24.1 GB · 128K 28.9 GB
Qwen3.8 27B
20.9 GB21.5 GB file
32K 22.2 GB · 128K 27.6 GB
BF16
Qwen3 30B A3B Instruct 2507
61.5 GB61.1 GB file
32K 63.0 GB · 128K 67.8 GB
Qwen3.8 27B
56.3 GB54.7 GB file
32K 57.7 GB · 128K 63.0 GB
Qwen3 30B A3B Instruct 2507
Ollama · Q4_K_M
ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_MQwen3.8 27B
Ollama · Q4_K_M
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:Q4_K_M| Spec | Qwen3 30B A3B Instruct 2507 | Qwen3.8 27B |
|---|---|---|
| Publisher | Alibaba | Alibaba |
| Parameters | 30.5B | 27.8B |
| Active parameters | 3.4B per token | — |
| Architecture | Qwen3MoeForCausalLM | Qwen3_5ForConditionalGeneration |
| Context | 256K tokens | 256K tokens |
| Modalities | Chat, Function-calling | Vision, Function-calling |
| License | Apache 2.0 | Apache 2.0 |
| Released | Jul 2025 | Aug 2026 |
Frequently Asked Questions
Which is better, Qwen3 30B A3B Instruct 2507 or Qwen3.8 27B?
Qwen3.8 27B leads llmrun Score 121.5 to 96.5; both fit 24 GB at Q4_K_M; Qwen3 30B A3B Instruct 2507 is 4.6× faster on RTX 4090 (~173 vs ~38 tok/s). Choose Qwen3 30B A3B Instruct 2507 for 4.6× faster on RTX 4090 and also does chat; choose Qwen3.8 27B for higher llmrun Score and stronger at reasoning.
Which needs less VRAM, Qwen3 30B A3B Instruct 2507 or Qwen3.8 27B?
At Q4_K_M, Qwen3 30B A3B Instruct 2507 requires 18.7 GB while Qwen3.8 27B needs 17.4 GB, making Qwen3.8 27B the easier model to run on your local setup.
Which is faster on RTX 4090?
Estimated decode speed on RTX 4090 is about 173 tok/s for Qwen3 30B A3B Instruct 2507 (Q4_K_M) and 38 tok/s for Qwen3.8 27B (Q4_K_M), so Qwen3 30B A3B Instruct 2507 is faster. Estimates, ±20%.
Can I run Qwen3 30B A3B Instruct 2507 or Qwen3.8 27B on a 16 GB GPU?
Qwen3 30B A3B Instruct 2507 fits at Q3_K_M; Qwen3.8 27B fits at Q3_K_L.
Which has a longer context window, Qwen3 30B A3B Instruct 2507 or Qwen3.8 27B?
Qwen3 30B A3B Instruct 2507 supports 262,144 tokens and Qwen3.8 27B supports 262,144 tokens.