llmrun

GPT OSS 20B vs Qwen3 30B A3B Thinking 2507

Qwen3 30B A3B Thinking 2507 leads llmrun Score 100.6 to 96.2 (ranges overlap); both fit 24 GB at Q4_K_M.

Model A · OpenAI

GPT OSS 20B
96.2llmrun Score

20.9B · 4.2B active · 128K context · Apache 2.0 · Aug 2025

ChatFunctions

Model B · Alibaba

Qwen3 30B A3B Thinking 2507
100.6llmrun Score

30.5B · 3.4B active · 256K context · Apache 2.0 · Jul 2025

ChatFunctions

Which should you choose?

Choose GPT OSS 20B if you want…

  • Fits a 12 GB GPU

    IQ4_XS · 11.8 GB; Qwen3 30B A3B Thinking 2507 needs 9.5 GB at its smallest quant

  • Lighter on memory

    12.9 vs 18.7 GB at Q4_K_M

Choose Qwen3 30B A3B Thinking 2507 if you want…

  • Higher llmrun Score

    100.6 vs 96.2, ranges overlap

  • Stronger at science

    98.8 vs 92.4

  • Ahead on the benchmarks

    6 of 8 shared boards

  • Longer context window

    256K vs 128K tokens

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

GPT OSS 20B

IQ2_M · ~103 tok/s

7.6 GB · up to 32Kfits

Qwen3 30B A3B Thinking 2507

Doesn't fit

smallest quant needs 9.5 GB

12 GB

RTX 3060 12GB · 12 GB tier

GPT OSS 20B

IQ4_XS · ~89 tok/s

11.8 GB · up to 16Kbetter quant

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

GPT OSS 20B

Q4_K_M · ~98 tok/s

13.1 GB · up to 128Kmore headroom

Qwen3 30B A3B Thinking 2507

Q3_K_M · ~119 tok/s

16 GB · up to 16K

24 GBreference

RTX 4090 · 24 GB tier

GPT OSS 20B

Q4_K_M · ~190 tok/s

13.1 GB · up to 128K

Qwen3 30B A3B Thinking 2507

Q4_K_M · ~173 tok/s

19.4 GB · up to 64K

32 GB

RTX 5090 · 32 GB tier

GPT OSS 20B

Q4_K_M · ~282 tok/s

13.1 GB · up to 128Kfaster

Qwen3 30B A3B Thinking 2507

Q4_K_M · ~227 tok/s

19.4 GB · up to 256K

48 GB

RTX A6000 · 48 GB tier

GPT OSS 20B

Q4_K_M · ~154 tok/s

13.1 GB · up to 128K

Qwen3 30B A3B Thinking 2507

Q4_K_M · ~148 tok/s

19.4 GB · up to 256K

Mac 32 GB

Mac Mini M4 · Macs

GPT OSS 20B

Q4_K_M · ~28 tok/s

13.1 GB · up to 128K

Qwen3 30B A3B Thinking 2507

Q4_K_M · ~29 tok/s

19.4 GB · up to 128K

Mac 64 GB

Mac Studio M4 Max · Macs

GPT OSS 20B

Q4_K_M · ~86 tok/s

13.1 GB · up to 128Kfaster

Qwen3 30B A3B Thinking 2507

Q4_K_M · ~67 tok/s

19.4 GB · up to 256K

Mac 128 GB

Mac Studio M4 Max · Macs

GPT OSS 20B

Q4_K_M · ~86 tok/s

13.1 GB · up to 128Kfaster

Qwen3 30B A3B Thinking 2507

Q4_K_M · ~67 tok/s

19.4 GB · up to 256K

Reference GPU: RTX 4090. Tinted cell = the better pick on that machine (fits, more headroom, or at least 10% faster).

llmrun Score

with its 90% range · bars share one scale

Overall

GPT OSS 20B

96.293–99

Qwen3 30B A3B Thinking 2507

100.697–104

Coding

GPT OSS 20B

100.992–110

Qwen3 30B A3B Thinking 2507

no data

Science

GPT OSS 20B

92.485–100

Qwen3 30B A3B Thinking 2507

98.891–106

Reasoning

GPT OSS 20B

90.882–99

Qwen3 30B A3B Thinking 2507

93.685–102

GPT OSS 20B · #58 of 92 open models · #15 of 34 that fit 24 GB

Qwen3 30B A3B Thinking 2507 · #45 of 92 open models · #9 of 34 that fit 24 GB

Benchmarks

8 shared boards: GPT OSS 20B leads 2, Qwen3 30B A3B Thinking 2507 leads 6 · Benchmarks imported Oct 2026

GPQA Diamond

GPT OSS 20B

60.8%#41/97 open

IndependentEpoch AI · Aug 2025

Qwen3 30B A3B Thinking 2507

70.1%#32/97 open

IndependentEpoch AI · Jul 2025

AIME 2024/2025

GPT OSS 20B

65.3%#33/85 open

IndependentEpoch AI · Aug 2025

Qwen3 30B A3B Thinking 2507

70.3%#27/85 open

IndependentEpoch AI · Jul 2025

MMLU-Pro

GPT OSS 20B

73.6%#33/144 open

IndependentTIGER-Lab

Qwen3 30B A3B Thinking 2507

80.9%#22/144 open

IndependentTIGER-Lab

SciCode

GPT OSS 20B

34.4%#47/65 open

Lab-reportedEpoch AI · Aug 2025

Qwen3 30B A3B Thinking 2507

33.3%#48/65 open

Lab-reportedEpoch AI · Jul 2025

Chess Puzzles

GPT OSS 20B

4.0%#33/67 open

IndependentEpoch AI · Aug 2025

Qwen3 30B A3B Thinking 2507

8.0%#26/67 open

IndependentEpoch AI · Jul 2025

CritPt

GPT OSS 20B

1.4%#30/65 open

Lab-reportedEpoch AI · Aug 2025

Qwen3 30B A3B Thinking 2507

0.3%#41/65 open

Lab-reportedEpoch AI · Jul 2025

DTBench

GPT OSS 20B

68.0%#34/68 open

Lab-reportedEpoch AI · Aug 2025

Qwen3 30B A3B Thinking 2507

69.3%#33/68 open

Lab-reportedEpoch AI · Jul 2025

LMCA

GPT OSS 20B

14.5%#43/54 open

Lab-reportedEpoch AI · Aug 2025

Qwen3 30B A3B Thinking 2507

19.5%#34/54 open

Lab-reportedEpoch AI · Jul 2025

4 more boards (other shared and single-model boards)

Only GPT OSS 20B (4)

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

GPT OSS 20B

9.3 GB11.5 GB file

32K 9.8 GB · 128K 11.5 GB

Qwen3 30B A3B Thinking 2507

13.4 GB11.3 GB file

32K 14.9 GB · 128K 19.7 GB

Q3_K_S

GPT OSS 20B

9.5 GB11.5 GB file

32K 10.0 GB · 128K 11.7 GB

Qwen3 30B A3B Thinking 2507

13.8 GB13.3 GB file

32K 15.3 GB · 128K 20.1 GB

Q3_K_M

GPT OSS 20B

10.6 GB11.5 GB file

32K 11.1 GB · 128K 12.8 GB

Qwen3 30B A3B Thinking 2507

15.3 GB14.7 GB file

32K 16.8 GB · 128K 21.6 GB

Q4_0

GPT OSS 20B

10.8 GB11.5 GB file

32K 11.3 GB · 128K 13.0 GB

Qwen3 30B A3B Thinking 2507

15.7 GB17.4 GB file

32K 17.2 GB · 128K 22.0 GB

Q4_K_M

GPT OSS 20B

12.9 GB11.6 GB file

32K 13.4 GB · 128K 15.1 GB

Qwen3 30B A3B Thinking 2507

18.7 GB18.6 GB file

32K 20.2 GB · 128K 25.1 GB

Q5_K_M

GPT OSS 20B

15.3 GB11.7 GB file

32K 15.8 GB · 128K 17.5 GB

Qwen3 30B A3B Thinking 2507

22.2 GB21.7 GB file

32K 23.7 GB · 128K 28.5 GB

Q6_K

GPT OSS 20B

17.6 GB12.0 GB file

32K 18.1 GB · 128K 19.8 GB

Qwen3 30B A3B Thinking 2507

25.6 GB25.1 GB file

32K 27.1 GB · 128K 31.9 GB

Q8_0

GPT OSS 20B

21.3 GB~20.9 GB file

32K 21.8 GB · 128K 23.5 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

GPT OSS 20B

6.1 GB~5.8 GB file

32K 6.6 GB · 128K 8.3 GB

Qwen3 30B A3B Thinking 2507

8.8 GB10.3 GB file

32K 10.3 GB · 128K 15.1 GB

IQ2_XS

GPT OSS 20B

6.7 GB~6.3 GB file

32K 7.2 GB · 128K 8.9 GB

Qwen3 30B A3B Thinking 2507

9.6 GB8.7 GB file

32K 11.1 GB · 128K 15.9 GB

IQ2_S

GPT OSS 20B

6.9 GB~6.5 GB file

32K 7.4 GB · 128K 9.1 GB

Qwen3 30B A3B Thinking 2507

9.9 GB8.7 GB file

32K 11.5 GB · 128K 16.3 GB

IQ2_M

GPT OSS 20B

7.4 GB~7.1 GB file

32K 7.9 GB · 128K 9.6 GB

Qwen3 30B A3B Thinking 2507

10.7 GB10.8 GB file

32K 12.2 GB · 128K 17.1 GB

IQ3_XXS

GPT OSS 20B

8.5 GB11.6 GB file

32K 9.0 GB · 128K 10.7 GB

Qwen3 30B A3B Thinking 2507

12.2 GB12.9 GB file

32K 13.7 GB · 128K 18.6 GB

Q2_K_S

GPT OSS 20B

8.7 GB12.1 GB file

32K 9.2 GB · 128K 10.9 GB

Qwen3 30B A3B Thinking 2507

12.6 GB10.5 GB file

32K 14.1 GB · 128K 19.0 GB

IQ3_XS

GPT OSS 20B

9.0 GB11.6 GB file

32K 9.5 GB · 128K 11.2 GB

Qwen3 30B A3B Thinking 2507

13.0 GB12.7 GB file

32K 14.5 GB · 128K 19.3 GB

IQ3_S

GPT OSS 20B

9.3 GB12.1 GB file

32K 9.8 GB · 128K 11.5 GB

Qwen3 30B A3B Thinking 2507

13.4 GB13.3 GB file

32K 14.9 GB · 128K 19.7 GB

IQ3_M

GPT OSS 20B

9.8 GB11.6 GB file

32K 10.3 GB · 128K 12.0 GB

Qwen3 30B A3B Thinking 2507

14.1 GB14.1 GB file

32K 15.7 GB · 128K 20.5 GB

Q3_K_L

GPT OSS 20B

11.1 GB11.5 GB file

32K 11.6 GB · 128K 13.3 GB

Qwen3 30B A3B Thinking 2507

16.1 GB14.6 GB file

32K 17.6 GB · 128K 22.4 GB

IQ4_XS

GPT OSS 20B

11.6 GB11.6 GB file

32K 12.1 GB · 128K 13.8 GB

Qwen3 30B A3B Thinking 2507

16.8 GB16.4 GB file

32K 18.3 GB · 128K 23.2 GB

IQ4_NL

GPT OSS 20B

12.1 GB11.6 GB file

32K 12.6 GB · 128K 14.3 GB

Qwen3 30B A3B Thinking 2507

17.6 GB17.3 GB file

32K 19.1 GB · 128K 23.9 GB

Q4_1

GPT OSS 20B

12.1 GB11.6 GB file

32K 12.6 GB · 128K 14.3 GB

Qwen3 30B A3B Thinking 2507

17.6 GB19.2 GB file

32K 19.1 GB · 128K 23.9 GB

Q4_K_S

GPT OSS 20B

12.1 GB11.6 GB file

32K 12.6 GB · 128K 14.3 GB

Qwen3 30B A3B Thinking 2507

17.6 GB17.5 GB file

32K 19.1 GB · 128K 23.9 GB

Q4_K_L

GPT OSS 20B

13.2 GB11.9 GB file

32K 13.7 GB · 128K 15.4 GB

Qwen3 30B A3B Thinking 2507

19.1 GB18.9 GB file

32K 20.6 GB · 128K 25.4 GB

Q5_0

GPT OSS 20B

not listed

Qwen3 30B A3B Thinking 2507

19.5 GB21.1 GB file

32K 21.0 GB · 128K 25.8 GB

Q5_K_S

GPT OSS 20B

14.8 GB11.7 GB file

32K 15.3 GB · 128K 17.0 GB

Qwen3 30B A3B Thinking 2507

21.4 GB21.1 GB file

32K 22.9 GB · 128K 27.7 GB

Q5_K_L

GPT OSS 20B

15.5 GB11.9 GB file

32K 16.0 GB · 128K 17.7 GB

Qwen3 30B A3B Thinking 2507

22.5 GB21.9 GB file

32K 24.1 GB · 128K 28.9 GB

BF16

GPT OSS 20B

42.2 GB~41.8 GB file

32K 42.7 GB · 128K 44.4 GB

Qwen3 30B A3B Thinking 2507

61.5 GB61.1 GB file

32K 63.0 GB · 128K 67.8 GB

How to run each

Specifications

Specifications of GPT OSS 20B and Qwen3 30B A3B Thinking 2507
SpecGPT OSS 20BQwen3 30B A3B Thinking 2507
PublisherOpenAIAlibaba
Parameters20.9B30.5B
Active parameters4.2B per token3.4B per token
ArchitectureGptOssForCausalLMQwen3MoeForCausalLM
Context128K tokens256K tokens
ModalitiesChat, Function-callingChat, Function-calling
LicenseApache 2.0Apache 2.0
ReleasedAug 2025Jul 2025

Frequently Asked Questions

Which is better, GPT OSS 20B or Qwen3 30B A3B Thinking 2507?

Qwen3 30B A3B Thinking 2507 leads llmrun Score 100.6 to 96.2 (ranges overlap); both fit 24 GB at Q4_K_M. Choose GPT OSS 20B for fits a 12 GB GPU and lighter on memory; choose Qwen3 30B A3B Thinking 2507 for higher llmrun Score and stronger at science.

Which needs less VRAM, GPT OSS 20B or Qwen3 30B A3B Thinking 2507?

At Q4_K_M, GPT OSS 20B requires 12.9 GB while Qwen3 30B A3B Thinking 2507 needs 18.7 GB, making GPT OSS 20B the easier model to run on your local setup.

Which is faster on RTX 4090?

Estimated decode speed on RTX 4090 is about 190 tok/s for GPT OSS 20B (Q4_K_M) and 173 tok/s for Qwen3 30B A3B Thinking 2507 (Q4_K_M), so GPT OSS 20B is faster. Estimates, ±20%.

Can I run GPT OSS 20B or Qwen3 30B A3B Thinking 2507 on a 16 GB GPU?

GPT OSS 20B fits at Q4_K_M; Qwen3 30B A3B Thinking 2507 fits at Q3_K_M.

Which has a longer context window, GPT OSS 20B or Qwen3 30B A3B Thinking 2507?

GPT OSS 20B supports 131,072 tokens and Qwen3 30B A3B Thinking 2507 supports 262,144 tokens.

GPT OSS 20B vs Qwen3 30B A3B Thinking 2507 — VRAM & GPU Comparison | llmrun