Qwen3 30B A3B Thinking 2507 — Benchmarks

Benchmark scores for Qwen3 30B A3B Thinking 2507 aggregated from public leaderboards (7 boards), with how it ranks among open models and how it compares with its nearest neighbours. See hardware requirements for what you need to run it.

llmrun Composite Score →

100.7

90% range 97.0–104.5 · from 7 boards

CodingNot enough data yet
AgentsNot enough data yet
MathNot enough data yet
Science98.790% range 91–106
Reasoning93.990% range 85–102

Qwen3 30B A3B Thinking 2507 reasoning benchmarks

Multi-step logic and problem-solving — can the model think through a hard problem rather than recall an answer.

GPQA Diamond
GPQA Diamond is a set of extremely hard, graduate-level science questions (physics, chemistry, biology) written by domain experts and filtered so that skilled non-experts with web access still fail. It measures genuine reasoning rather than memorization.
70.1%
70.1%
median 55.7%open leader 93.2%
Overall best: Gemini 4 Argon (high) 96.4%
66th percentile · #35 of 100 open · #183 of 309 all
#35 of 100
#183 of 309
Independent
Epoch AI · Jul 2025
CritPt
CritPt (Critical Points) tests graduate-level physics problem-solving on research-style questions spanning mechanics, electromagnetism, quantum mechanics, and more. It was created by physicists to probe whether a model can reason through multi-step physics rather than pattern-match textbook answers.
0.3%
0.3%
median 0.9%open leader 26.6%
Overall best: Gemini 4 Carbon (high) 36.0%
42nd percentile · #49 of 84 open · #189 of 244 all
#49 of 84
#189 of 244
Lab-reported
Epoch AI · Jul 2025
DTBench
DTBench is a set of handcrafted, expert-validated multiple-choice questions on decision theory — Newcomb-like problems and related puzzles — built by the Conceptual Reasoning Index project with Anthropic. It tests whether a model can follow an abstract argument rather than recall a fact.
69.3%
69.3%
median 68.7%open leader 93.9%
Overall best: Claude Opus 5.5 Max 98.9%
51st percentile · #35 of 70 open · #160 of 235 all
#35 of 70
#160 of 235
Lab-reported
Epoch AI · Jul 2025
LMCA
LMCA (Language Model Conceptual Argumentation) scores models on expert-rated arguments across philosophy, decision theory, and AI risk — domains where there is no empirical feedback loop to check an answer against. It is part of the Conceptual Reasoning Index, built with Anthropic by researchers Emery Cooper and Caspar Oesterheld.
19.5%
19.5%
median 25.5%open leader 55.5%
Overall best: Claude Sonnet 5.5 Max 68.3%
36th percentile · #36 of 56 open · #162 of 197 all
#36 of 56
#162 of 197
Lab-reported
Epoch AI · Jul 2025

Qwen3 30B A3B Thinking 2507 coding benchmarks

Writing, editing, and fixing real code — measured by whether the resulting program actually runs and passes tests.

SciCode
SciCode asks a model to write real scientific-computing code — implementing multi-step numerical methods drawn from physics, chemistry, biology, and math research — and checks the output against reference solutions. It was built by a team of scientists and ML researchers to test coding ability on genuine research problems, not typical software tasks.
33.0%
33.0%
median 40.1%open leader 60.9%
Overall best: Claude Opus 5.5 Max 66.9%
23rd percentile · #56 of 72 open · #196 of 217 all
#56 of 72
#196 of 217
Lab-reported
Epoch AI · Jul 2025

Qwen3 30B A3B Thinking 2507 math benchmarks

Symbolic and quantitative problem-solving, from competition math to multi-step calculation.

AIME 2024/2025
AIME (American Invitational Mathematics Examination) is a prestigious high-school competition of hard, integer-answer problems. It's a widely-cited yardstick for multi-step mathematical reasoning, here on the 2024–2025 papers.
70.3%
70.3%
median 54.4%open leader 99.4%
Overall best: Claude Fable 5 (high) 100.0%
67th percentile · #30 of 88 open · #150 of 292 all
#30 of 88
#150 of 292
Independent
Epoch AI · Jul 2025

Qwen3 30B A3B Thinking 2507 knowledge benchmarks

Breadth and depth of factual knowledge across many subjects, paired with the reasoning to apply it.

MMLU-Pro
MMLU-Pro is a harder, cleaned-up successor to MMLU with ten answer choices and more reasoning-heavy questions across 14 subjects, measuring broad knowledge and reasoning together.
80.9%
80.9%
median 52.0%open leader 88.0%
Overall best: Gemini-3.1-Pro 91.2%
86th percentile · #22 of 146 open · #60 of 257 all
#22 of 146
#60 of 257
Independent
TIGER-Lab

Score vs VRAM in the 24 GB tier

Qwen3 30B A3B Thinking 2507
Other models
Best Score for its VRAM
x: VRAM (GB) · y: llmrun Score

#9 of 36 that fit 24 GB

Each dot is a model at the best quant that fits a 24 GB GPU. The line steps up to the best Score available at each amount of VRAM; a dot below it is beaten by a model that needs no more memory.

Full 24 GB ranking →

Qwen3 30B A3B Thinking 2507 vs its neighbours

Same boards for the nearest open models per llmrun Score and the best model that fits 24 GB. The highest score in each row is marked.

Board	Qwen3 30B A3B Thinking 2507this model	Qwen3.8 27Bbest in classQ4_K_M · 18 GB	Seed OSS 36B Instructnearest by ScoreQ4_K_S · 23 GB	Qwen3 32Bnearest by ScoreQ4_K_M · 23 GB	Qwen3 14Bnearest by ScoreQ4_K_M · 12 GB
llmrun Score	100.7	122.0	102.5	98.6	98.0
GPQA Diamond	70.1%	—	71.5%	65.7%	63.8%
CritPt	0.3%	5.4%	—	0.3%	0.0%
DTBench	69.3%	88.0%	—	67.5%	64.0%
LMCA	19.5%	41.4%	—	17.3%	18.2%
SciCode	33.0%	46.6%	—	35.4%	31.6%
AIME 2024/2025	70.3%	—	67.5%	66.9%	66.4%
MMLU-Pro	80.9%	—	82.7%	—	—
Compare head to head		Compare	Compare	Compare	Compare

Scores aggregated from public benchmark sources (Epoch AI: 6 boards, TIGER-Lab: 1 board) — each linked from the benchmark pages. llmrun does not run these benchmarks. Percentiles and ranks are among open models in the llmrun catalogue; the median and leader marks refer to the same field.