Llama 4 Scout 17B 16E Instruct vs Qwen3 235B A22B Thinking 2507

Qwen3 235B A22B Thinking 2507 leads llmrun Score 108.2 to 81.8; neither fits 24 GB.

Model A · Meta

Llama 4 Scout 17B 16E Instruct

81.8llmrun Score

108.6B · 17B active · Other · Apr 2025

Vision

Model B · Alibaba

Qwen3 235B A22B Thinking 2507

108.2llmrun Score

235.1B · 22.2B active · 256K context · Apache 2.0 · Jul 2025

ChatFunctions

Which should you choose?

Choose Llama 4 Scout 17B 16E Instruct if you want…

Fits a Mac 64 GB
IQ3_M · 59.1 GB; Qwen3 235B A22B Thinking 2507 needs 72.4 GB at its smallest quant
Lighter on memory
71.7 vs 141.6 GB at Q4_K_M
Also does vision
Qwen3 235B A22B Thinking 2507 doesn't list it

Choose Qwen3 235B A22B Thinking 2507 if you want…

Higher llmrun Score
108.2 vs 81.8
Stronger at reasoning
106.4 vs 79.7
Stronger at science
107.5 vs 83.0
Ahead on the benchmarks
7 of 8 shared boards
Also does chat, function-calling
Llama 4 Scout 17B 16E Instruct doesn't list it
More permissive license
Apache 2.0 vs Other

On your hardware

Each model at its default quant (Q4_K_M), a smaller one when it must. Estimates, ±20%.

MemoryLlama 4 Scout 17B 16E InstructQwen3 235B A22B Thinking 2507

8 GB

RTX 4060 · 8 GB tier

Llama 4 Scout 17B 16E Instruct

Doesn't fit

smallest quant needs 38.2 GB

Qwen3 235B A22B Thinking 2507

Doesn't fit

smallest quant needs 72.4 GB

12 GB

RTX 3060 12GB · 12 GB tier

Llama 4 Scout 17B 16E Instruct

Doesn't fit

smallest quant needs 38.2 GB

Qwen3 235B A22B Thinking 2507

Doesn't fit

smallest quant needs 72.4 GB

16 GB

RTX 5060 Ti 16GB · 16 GB tier

Llama 4 Scout 17B 16E Instruct

Doesn't fit

smallest quant needs 38.2 GB

Qwen3 235B A22B Thinking 2507

Doesn't fit

smallest quant needs 72.4 GB

24 GBreference

RTX 4090 · 24 GB tier

Llama 4 Scout 17B 16E Instruct

Doesn't fit

smallest quant needs 38.2 GB

Qwen3 235B A22B Thinking 2507

Doesn't fit

smallest quant needs 72.4 GB

32 GB

RTX 5090 · 32 GB tier

Llama 4 Scout 17B 16E Instruct

Doesn't fit

smallest quant needs 38.2 GB

Qwen3 235B A22B Thinking 2507

Doesn't fit

smallest quant needs 72.4 GB

48 GB

RTX A6000 · 48 GB tier

Llama 4 Scout 17B 16E Instruct

IQ2_M · ~65 tok/s

45.7 GB · up to 16Kfits

Qwen3 235B A22B Thinking 2507

Doesn't fit

smallest quant needs 72.4 GB

Mac 32 GB

Mac Mini M4 · Macs

Llama 4 Scout 17B 16E Instruct

Doesn't fit

smallest quant needs 38.2 GB

Qwen3 235B A22B Thinking 2507

Doesn't fit

smallest quant needs 72.4 GB

Mac 64 GB

Mac Studio M4 Max · Macs

Llama 4 Scout 17B 16E Instruct

IQ3_M · ~32 tok/s

59.1 GB · up to 16Kfits

Qwen3 235B A22B Thinking 2507

Doesn't fit

smallest quant needs 72.4 GB

Mac 128 GB

Mac Studio M4 Max · Macs

Llama 4 Scout 17B 16E Instruct

Q4_K_M · ~26 tok/s

77 GB · up to 32Kmore headroom

Qwen3 235B A22B Thinking 2507

Q3_K_L · ~21 tok/s

122.4 GB · up to 32K

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

Llama 4 Scout 17B 16E Instruct Qwen3 235B A22B Thinking 2507

Overall

Llama 4 Scout 17B 16E Instruct

81.878–86

Qwen3 235B A22B Thinking 2507

108.2105–112

Coding

Llama 4 Scout 17B 16E Instruct

no data

Qwen3 235B A22B Thinking 2507

102.392–112

Science

Llama 4 Scout 17B 16E Instruct

83.075–91

Qwen3 235B A22B Thinking 2507

107.599–116

Reasoning

Llama 4 Scout 17B 16E Instruct

79.770–89

Qwen3 235B A22B Thinking 2507

106.499–114

Llama 4 Scout 17B 16E Instruct · #79 of 100 open models · #40 of 60 that fit ≤96 GB

Qwen3 235B A22B Thinking 2507 · #33 of 100 open models · #16 of 78 that fit ≤192 GB

Benchmarks

8 shared boards: Llama 4 Scout 17B 16E Instruct leads 0, Qwen3 235B A22B Thinking 2507 leads 7 · Benchmarks imported Oct 2026

BoardLlama 4 Scout 17B 16E InstructQwen3 235B A22B Thinking 2507

GPQA Diamond

Llama 4 Scout 17B 16E Instruct

51.8%#51/98 open

IndependentEpoch AI · Apr 2025

Qwen3 235B A22B Thinking 2507

80.0%#25/98 open

IndependentEpoch AI · Jul 2025

AIME 2024/2025

Llama 4 Scout 17B 16E Instruct

7.8%#62/86 open

IndependentEpoch AI · Apr 2025

Qwen3 235B A22B Thinking 2507

86.7%#19/86 open

IndependentEpoch AI · Jul 2025

MMLU-Pro

Llama 4 Scout 17B 16E Instruct

74.3%#31/145 open

IndependentTIGER-Lab

Qwen3 235B A22B Thinking 2507

84.5%#11/145 open

IndependentTIGER-Lab

SciCode

Llama 4 Scout 17B 16E Instruct

17.0%#63/66 open

Lab-reportedEpoch AI · Apr 2025

Qwen3 235B A22B Thinking 2507

41.4%#27/66 open

Lab-reportedEpoch AI · Jul 2025

CritPt

Llama 4 Scout 17B 16E Instruct

0.0%#56/66 open

Lab-reportedEpoch AI · Apr 2025

Qwen3 235B A22B Thinking 2507

0.0%#63/66 open

Lab-reportedEpoch AI · Jul 2025

DTBench

Llama 4 Scout 17B 16E Instruct

57.9%#53/68 open

Lab-reportedEpoch AI · Apr 2025

Qwen3 235B A22B Thinking 2507

80.3%#22/68 open

Lab-reportedEpoch AI · Jul 2025

Fiction.LiveBench

Llama 4 Scout 17B 16E Instruct

36.0%#22/25 open

IndependentEpoch AI · Apr 2025

Qwen3 235B A22B Thinking 2507

75.0%#5/25 open

IndependentEpoch AI · Jul 2025

LMCA

Llama 4 Scout 17B 16E Instruct

12.0%#46/54 open

Lab-reportedEpoch AI · Apr 2025

Qwen3 235B A22B Thinking 2507

29.3%#23/54 open

Lab-reportedEpoch AI · Jul 2025

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

LMArena Text

Llama 4 Scout 17B 16E Instruct

1321.5#94/187 open

ArenaLMArena · Oct 2026

Qwen3 235B A22B Thinking 2507

1399.8#49/187 open

ArenaLMArena · Oct 2026

Only Llama 4 Scout 17B 16E Instruct (4)

Only Qwen3 235B A22B Thinking 2507 (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.

QuantLlama 4 Scout 17B 16E InstructQwen3 235B A22B Thinking 2507

Q2_K

Llama 4 Scout 17B 16E Instruct

50.8 GB39.6 GB file

32K 62.2 GB · 128K 98.7 GB

Qwen3 235B A22B Thinking 2507

100 GB85.7 GB file

32K 103 GB · 128K 113 GB

Q3_K_S

Llama 4 Scout 17B 16E Instruct

52.3 GB46.7 GB file

32K 63.7 GB · 128K 100 GB

Qwen3 235B A22B Thinking 2507

103 GB101.4 GB file

32K 106 GB · 128K 116 GB

Q3_K_M

Llama 4 Scout 17B 16E Instruct

58.3 GB51.8 GB file

32K 69.7 GB · 128K 106 GB

Qwen3 235B A22B Thinking 2507

115 GB112.4 GB file

32K 118 GB · 128K 128 GB

Q4_0

Llama 4 Scout 17B 16E Instruct

59.8 GB61.2 GB file

32K 71.2 GB · 128K 108 GB

Qwen3 235B A22B Thinking 2507

118 GB133.1 GB file

32K 121 GB · 128K 130 GB

Q4_K_M

Llama 4 Scout 17B 16E Instruct

71.7 GB65.4 GB file

32K 83.1 GB · 128K 120 GB

Qwen3 235B A22B Thinking 2507

142 GB142.2 GB file

32K 145 GB · 128K 154 GB

Q5_K_M

Llama 4 Scout 17B 16E Instruct

85.2 GB76.5 GB file

32K 96.6 GB · 128K 133 GB

Qwen3 235B A22B Thinking 2507

168 GB166.8 GB file

32K 171 GB · 128K 180 GB

Q6_K

Llama 4 Scout 17B 16E Instruct

98.6 GB88.4 GB file

32K 110 GB · 128K 147 GB

Qwen3 235B A22B Thinking 2507

194 GB193.0 GB file

32K 197 GB · 128K 207 GB

Q8_0

Llama 4 Scout 17B 16E Instruct

120 GB114.5 GB file

32K 131 GB · 128K 167 GB

Qwen3 235B A22B Thinking 2507

236 GB249.9 GB file

32K 239 GB · 128K 248 GB

16 more quants

IQ2_XXS

Llama 4 Scout 17B 16E Instruct

32.9 GB37.4 GB file

32K 44.3 GB · 128K 80.8 GB

Qwen3 235B A22B Thinking 2507

not listed

IQ2_XS

Llama 4 Scout 17B 16E Instruct

35.9 GB31.8 GB file

32K 47.3 GB · 128K 83.8 GB

Qwen3 235B A22B Thinking 2507

71.0 GB65.6 GB file

32K 74.0 GB · 128K 83.5 GB

IQ2_S

Llama 4 Scout 17B 16E Instruct

37.4 GB32.3 GB file

32K 48.8 GB · 128K 85.3 GB

Qwen3 235B A22B Thinking 2507

74.0 GB66.1 GB file

32K 76.9 GB · 128K 86.4 GB

IQ2_M

Llama 4 Scout 17B 16E Instruct

40.3 GB39.1 GB file

32K 51.8 GB · 128K 88.3 GB

Qwen3 235B A22B Thinking 2507

79.8 GB78.6 GB file

32K 82.8 GB · 128K 92.3 GB

IQ3_XXS

Llama 4 Scout 17B 16E Instruct

46.3 GB45.7 GB file

32K 57.7 GB · 128K 94.3 GB

Qwen3 235B A22B Thinking 2507

91.6 GB93.1 GB file

32K 94.6 GB · 128K 104 GB

Q2_K_S

Llama 4 Scout 17B 16E Instruct

47.8 GB37.0 GB file

32K 59.2 GB · 128K 95.8 GB

Qwen3 235B A22B Thinking 2507

not listed

IQ3_XS

Llama 4 Scout 17B 16E Instruct

49.3 GB44.3 GB file

32K 60.7 GB · 128K 97.3 GB

Qwen3 235B A22B Thinking 2507

97.5 GB96.9 GB file

32K 100 GB · 128K 110 GB

IQ3_S

Llama 4 Scout 17B 16E Instruct

50.8 GB46.8 GB file

32K 62.2 GB · 128K 98.7 GB

Qwen3 235B A22B Thinking 2507

not listed

IQ3_M

Llama 4 Scout 17B 16E Instruct

53.8 GB47.5 GB file

32K 65.2 GB · 128K 102 GB

Qwen3 235B A22B Thinking 2507

106 GB107.3 GB file

32K 109 GB · 128K 119 GB

Q3_K_L

Llama 4 Scout 17B 16E Instruct

61.3 GB57.8 GB file

32K 72.7 GB · 128K 109 GB

Qwen3 235B A22B Thinking 2507

121 GB111.2 GB file

32K 124 GB · 128K 133 GB

IQ4_XS

Llama 4 Scout 17B 16E Instruct

64.2 GB57.7 GB file

32K 75.7 GB · 128K 112 GB

Qwen3 235B A22B Thinking 2507

127 GB125.5 GB file

32K 130 GB · 128K 139 GB

IQ4_NL

Llama 4 Scout 17B 16E Instruct

67.2 GB60.9 GB file

32K 78.6 GB · 128K 115 GB

Qwen3 235B A22B Thinking 2507

133 GB133.1 GB file

32K 136 GB · 128K 145 GB

Q4_1

Llama 4 Scout 17B 16E Instruct

67.2 GB67.6 GB file

32K 78.6 GB · 128K 115 GB

Qwen3 235B A22B Thinking 2507

133 GB147.2 GB file

32K 136 GB · 128K 145 GB

Q4_K_S

Llama 4 Scout 17B 16E Instruct

67.2 GB61.5 GB file

32K 78.6 GB · 128K 115 GB

Qwen3 235B A22B Thinking 2507

133 GB133.7 GB file

32K 136 GB · 128K 145 GB

Q5_K_S

Llama 4 Scout 17B 16E Instruct

82.2 GB74.3 GB file

32K 93.6 GB · 128K 130 GB

Qwen3 235B A22B Thinking 2507

162 GB161.9 GB file

32K 165 GB · 128K 175 GB

BF16

Llama 4 Scout 17B 16E Instruct

239 GB215.6 GB file

32K 250 GB · 128K 287 GB

Qwen3 235B A22B Thinking 2507

471 GB470.3 GB file

32K 474 GB · 128K 483 GB

How to run each

Llama 4 Scout 17B 16E Instruct

Ollama · Q4_K_M

$ollama run hf.co/unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF:Q4_K_M

All ways to run Llama 4 Scout 17B 16E Instruct →

Qwen3 235B A22B Thinking 2507

Ollama · Q4_K_M

$ollama run hf.co/unsloth/Qwen3-235B-A22B-Thinking-2507-GGUF:Q4_K_M

All ways to run Qwen3 235B A22B Thinking 2507 →

Specifications

Specifications of Llama 4 Scout 17B 16E Instruct and Qwen3 235B A22B Thinking 2507
Spec	Llama 4 Scout 17B 16E Instruct	Qwen3 235B A22B Thinking 2507
Publisher	Meta	Alibaba
Parameters	108.6B	235.1B
Active parameters	17B per token	22B per token
Architecture	—	Qwen3MoeForCausalLM
Context	—	256K tokens
Modalities	Vision	Chat, Function-calling
License	Other	Apache 2.0
Released	Apr 2025	Jul 2025

Frequently Asked Questions

Which is better, Llama 4 Scout 17B 16E Instruct or Qwen3 235B A22B Thinking 2507?

Qwen3 235B A22B Thinking 2507 leads llmrun Score 108.2 to 81.8; neither fits 24 GB. Choose Llama 4 Scout 17B 16E Instruct for fits a Mac 64 GB and lighter on memory; choose Qwen3 235B A22B Thinking 2507 for higher llmrun Score and stronger at reasoning.

Which needs less VRAM, Llama 4 Scout 17B 16E Instruct or Qwen3 235B A22B Thinking 2507?

At Q4_K_M, Llama 4 Scout 17B 16E Instruct requires 71.7 GB while Qwen3 235B A22B Thinking 2507 needs 141.6 GB, making Llama 4 Scout 17B 16E Instruct the easier model to run on your local setup.

Can I run Llama 4 Scout 17B 16E Instruct or Qwen3 235B A22B Thinking 2507 on a 16 GB GPU?

Llama 4 Scout 17B 16E Instruct does not fit (smallest quant needs 38.2 GB); Qwen3 235B A22B Thinking 2507 does not fit (smallest quant needs 72.4 GB).