GeekRadar / The comparison desk

COMPARE.

AI models

alibaba

Qwen3 235B-A22B Instruct 2507

AI model Released Jul 21, 2025

Context window: 262,144 tokens
Maximum output: 16,384 tokens
Total parameters: 235.09 B
Open weights: Yes

Capability & preference

Human preference · Arena

Overall text preferences from head-to-head votes. Not factual accuracy or a task-success rate.

Qwen3 235B-A22B Instruct 2507 1419.1 qwen3-235b-a22b-instruct-2507 · 95% interval 1416.6–1421.7 · 98,134 votes · 2026-10-15

Recorded results

Only the same benchmark, version, workload, scenario and unit share a row. Missing results are not zero. System results include the full submitted hardware and software configuration.

Arena text preference / chinesehuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,464.63 rating

Confidence interval: 1,457.40 – 1,471.87

7309 votes

Oct 15, 2026

Category: chinese
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / codinghuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,444.51 rating

Confidence interval: 1,440.04 – 1,448.99

21747 votes

Oct 15, 2026

Category: coding
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / creative_writinghuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,374.80 rating

Confidence interval: 1,369.42 – 1,380.19

13895 votes

Oct 15, 2026

Category: creative_writing
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / englishhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,423.85 rating

Confidence interval: 1,420.34 – 1,427.35

41055 votes

Oct 15, 2026

Category: english
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / exclude_tieshuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,407.68 rating

Confidence interval: 1,404.15 – 1,411.20

69844 votes

Oct 15, 2026

Category: exclude_ties
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / experthuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,441.76 rating

Confidence interval: 1,434.26 – 1,449.26

6453 votes

Oct 15, 2026

Category: expert
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / frenchhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,444.39 rating

Confidence interval: 1,430.87 – 1,457.90

2166 votes

Oct 15, 2026

Category: french
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / germanhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,432.17 rating

Confidence interval: 1,417.97 – 1,446.37

1812 votes

Oct 15, 2026

Category: german
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / hard_promptshuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,433.34 rating

Confidence interval: 1,430.09 – 1,436.58

53546 votes

Oct 15, 2026

Category: hard_prompts
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / hard_prompts_englishhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,436.09 rating

Confidence interval: 1,431.68 – 1,440.51

22730 votes

Oct 15, 2026

Category: hard_prompts_english
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,418.06 rating

Confidence interval: 1,413.23 – 1,422.89

18259 votes

Oct 15, 2026

Category: industry_business_and_management_and_financial_operations
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,374.17 rating

Confidence interval: 1,369.37 – 1,378.97

18057 votes

Oct 15, 2026

Category: industry_entertainment_and_sports_and_media
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,428.32 rating

Confidence interval: 1,420.83 – 1,435.81

6607 votes

Oct 15, 2026

Category: industry_legal_and_government
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,437.33 rating

Confidence interval: 1,432.25 – 1,442.42

15760 votes

Oct 15, 2026

Category: industry_life_and_physical_and_social_science
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / industry_mathematicalhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,440.54 rating

Confidence interval: 1,432.27 – 1,448.82

5215 votes

Oct 15, 2026

Category: industry_mathematical
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,444.99 rating

Confidence interval: 1,437.06 – 1,452.93

5922 votes

Oct 15, 2026

Category: industry_medicine_and_healthcare
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,446.50 rating

Confidence interval: 1,442.75 – 1,450.26

34910 votes

Oct 15, 2026

Category: industry_software_and_it_services
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,390.06 rating

Confidence interval: 1,385.63 – 1,394.49

22136 votes

Oct 15, 2026

Category: industry_writing_and_literature_and_language
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / instruction_followinghuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,407.80 rating

Confidence interval: 1,403.75 – 1,411.86

27846 votes

Oct 15, 2026

Category: instruction_following
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / japanesehuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,399.02 rating

Confidence interval: 1,382.83 – 1,415.21

1439 votes

Oct 15, 2026

Category: japanese
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / koreanhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,390.57 rating

Confidence interval: 1,375.75 – 1,405.39

1821 votes

Oct 15, 2026

Category: korean
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / longer_queryhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,425.99 rating

Confidence interval: 1,421.82 – 1,430.16

27123 votes

Oct 15, 2026

Category: longer_query
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / mathhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,431.53 rating

Confidence interval: 1,423.89 – 1,439.17

6077 votes

Oct 15, 2026

Category: math
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / multi_turnhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,432.10 rating

Confidence interval: 1,427.10 – 1,437.09

17054 votes

Oct 15, 2026

Category: multi_turn
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / non_englishhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,409.36 rating

Confidence interval: 1,406.22 – 1,412.51

57028 votes

Oct 15, 2026

Category: non_english
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / overallhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,419.13 rating

Confidence interval: 1,416.59 – 1,421.66

98134 votes

Oct 15, 2026

Category: overall
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / polishhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,406.63 rating

Confidence interval: 1,396.98 – 1,416.28

3945 votes

Oct 15, 2026

Category: polish
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / russianhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,410.86 rating

Confidence interval: 1,404.86 – 1,416.85

10181 votes

Oct 15, 2026

Category: russian
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Arena text preference / spanishhuman preference · 2026-10-15 · rating

Qwen3 235B-A22B Instruct 2507

1,429.53 rating

Confidence interval: 1,417.85 – 1,441.21

2911 votes

Oct 15, 2026

Category: spanish
Evaluated configuration: qwen3-235b-a22b-instruct-2507
Evaluation type: human preference

Run it locally

Inference · memory estimate

Qwen3 235B-A22B Instruct 2507

235.09 B total parameters · qwen3_moe · BF16

Published checkpoint tensor payload: 437.9 GiB. This includes stored copies and buffers; it is separate from the runtime estimates below.

Weight precision Total context Check a GPU Check a Local AI system

Estimated accelerator memory133.2 GiB

Weights: 120.4 GiB
KV cache: 0.7 GiB
Runtime allowance: 12.0 GiB

One user, full model in accelerator memory.

GPUs with memory headroom

Local AI systems with memory headroom

AMD Ryzen AI Halo (Ryzen AI Max+ PRO 495, 192 GB) Planning budget: 144.0 GiB
GMKtec EVO-X5 Pro (Ryzen AI Max+ PRO 495, 192 GB, 2 TB) Planning budget: 144.0 GiB
Minisforum MS-S1 MAX-P495 (192 GB, 2 TB) Planning budget: 144.0 GiB
Apple Mac Studio (M5 Ultra 36-core CPU, 80-core GPU, 256 GB) Planning budget: 192.0 GiB
Apple Mac Studio (M5 Ultra 36-core CPU, 80-core GPU, 512 GB) Planning budget: 384.0 GiB

Shared-memory systems use a planning budget of 75% of installed memory, leaving 25% for the OS and other allocations. This is an assumption, not a measured GPU allocation. Firmware and driver limits may be lower and may require configuration. Memory matches do not verify a runtime, quantized model artifact or inference speed.

Includes FP16 KV cache, prompt plus generated tokens, quantization metadata and a runtime allowance. GPU matches reserve 10% of VRAM. Memory fit is not a speed or software-compatibility guarantee; quantized weights and runtime support must be available.

API cost

USD / month

Input tokens / monthOutput tokens / month

Standard text-token estimate. Excludes taxes, caching, batch discounts, reasoning and multimodal surcharges. Provider limits and regional availability may differ.

Qwen3 235B-A22B Instruct 2507

Provider offer

Qwen3 235B-A22B Instruct 2507 / Pricing tiers 1

Provider / scope	Rate	Currency	Amount / unit
DeepInfraQwen/Qwen3-235B-A22B-Instruct-2507Standard	Input	USD	0.09 / 1M tokens
DeepInfraQwen/Qwen3-235B-A22B-Instruct-2507Standard	Output	USD	0.55 / 1M tokens

Capabilities

Context window: 262,144 tokens
Maximum output: 16,384 tokens
Reasoning: No
Tool calling: Yes
Open weights: Yes
Model license: Apache 2.0
Total parameters: 235.09 B
Checkpoint tensor payload: 437.90 GiB

Input modalities

text

Output modalities

text

Processing & platform

Architecture: qwen3_moe