Skip to main content
Meo

Qwen: Qwen3 VL 30B A3B Instruct

Qwen: Qwen3 VL 30B A3B Instruct is qwen's model. On the un-leaked meo benchmark it ranks #31 of 82 with 66.1% accuracy and an Effective Value (𝕍) of 0.0053 β€” earning a correct answer for about $0.0023 in 64s each.

as of 2026-10-18

meo rank
#31
of 82
Accuracy
66.1%
overall
𝕍 index (top=100)
0.13
raw 𝕍 0.0053 Β· N=10
$ / correct
$0.0023
64s each

Per-domain accuracy

166 correct of 250 questions asked across 11 domains Β· answer rate 100% Β· macro accuracy 70.8%. Headline accuracy is micro-averaged per meo’s methodology; per-domain question counts (correct/asked) are shown so you can judge significance.

State tracking
100%1/1
CSP (unsatisfiable)
94%15/16
Regex automaton
88%14/16
Base-rate bias
83%34/41
Watson-Glaser
75%3/4
Tape machine
73%11/15
Logic / math / CS
71%24/34
Framework bias
67%2/3
Theory of mind
67%2/3
Perceptual illusions
54%55/102
Long arithmetic
33%5/15

Efficiency

Cost per correct
$0.0023
Seconds per correct
64
Tokens per correct
3,599
Correct answers
166 / 250
Total benchmark cost
$0.38
Avg tokens / item
2,390

Model details

Released
2025-10-06
Weights
Open
Context
262K
Max output
16K
Knowledge cutoff
2025-03-31
Price in
$0.15/M
Price out
$0.60/M
Modality
text+image->text
View on OpenRouter β†—Hugging Face β†—

Sources: openrouter Β· meo

FAQ

How accurate is Qwen: Qwen3 VL 30B A3B Instruct?

Qwen: Qwen3 VL 30B A3B Instruct scores 66.1% overall on the meo benchmark (rank #31 of 82), an un-leaked private holdout graded objective-first.

What is Qwen: Qwen3 VL 30B A3B Instruct's Effective Value?

Its Effective Value (𝕍) is 0.0053 at N=10, Ο‰=1, Ξ΄=1.5 β€” a fused metric of accuracy, speed, cost, and error-cascade. 𝕍 rankings shift with task depth.

How much does Qwen: Qwen3 VL 30B A3B Instruct cost per correct answer?

About $0.0023 per correct answer (3,599 tokens, 64 seconds each) at benchmark-time pricing via OpenRouter.

Is Qwen: Qwen3 VL 30B A3B Instruct open weights?

Yes β€” Qwen: Qwen3 VL 30B A3B Instruct is released with open weights.

Benchmark updates affecting this model

  • 2026-06-09majorlogic_math_cs ground truth was broken β€” culled and rebuilt with solver-verified items

    A 5-lab frontier jury revealed that 29 of 37 LLM-authored logic_math_cs items had broken ground truth: on most, five independent frontier models unanimously concluded the clue sets were contradictory while the stored answer claimed a number, and on eight, all five agreed on the SAME different number (author arithmetic errors). Models were being marked wrong for being right. The 29 were retired (full audit published in the repo) and the domain rebuilt with generator-as-oracle items: an embedded constraint solver PROVES each puzzle has exactly one solution (or provably none), verified by an independent exhaustive enumerator. A 3-round difficulty study found frontier models saturate consistent human-scale logic puzzles, so items use anonymous global constraints and carry a measured difficulty gradient.

  • 2026-06-08majorCorrected base_bias grading (LLM-equivalence judge fix)

    The LLM-equivalence grader uses a reasoning model whose verdict token budget was too small, so its YES/NO answer was emitted as empty and silently counted as a failure. This understated the base_bias domain to ~7% for every model when the true accuracy is ~85-90%. We raised the judge budget, excluded its reasoning tokens from the verdict, switched to verbosity-robust conclusion matching, and re-graded the whole board.

Full benchmark changelog β†’

Embed this ranking

Show Qwen: Qwen3 VL 30B A3B Instruct’s meo rank on your site β€” the badge links back here.

Related: all Qwen (Alibaba) models Β· Multi-step state tracking domain Β· best Effective Value Β· biggest movers Β· methodology Β· full leaderboard.