Skip to main content

Tencent: Hy3 preview

Tencent: Hy3 preview is tencent's model. On the un-leaked meo benchmark it ranks #21 of 22 with 25.0% accuracy and an Effective Value (𝕍) of 0.0000 β€” earning a correct answer for about $0.0140 in 579s each.

as of 2026-06-08

meo rank
#21
of 22
Accuracy
25.0%
overall
𝕍 index (top=100)
<0.001
raw 𝕍 0.0000 Β· N=10
$ / correct
$0.0140
579s each

Per-domain accuracy

Theory of mind
67%
CSP (unsatisfiable)
50%
Regex automaton
50%
Tape machine
33%
Framework bias
33%
Watson-Glaser
25%
Base-rate bias
7%
Long arithmetic
7%
Logic / math / CS
3%
State tracking
0%

Efficiency

Cost per correct
$0.0140
Seconds per correct
579
Tokens per correct
57,712
Correct answers
30 / 151
Total benchmark cost
$0.42
Avg tokens / item
11,466

Aggregated third-party benchmarks

Artificial Analysis (artificialanalysis.ai). Redistribution requires an AA commercial license.

Model details

Lab
tencent
Released
2026-04-22
Weights
Open
Context
262K
Max output
β€”
Knowledge cutoff
β€”
Price in
$0.06/M
Price out
$0.21/M
Modality
text->text

Sources: openrouter Β· artificial_analysis Β· meo

FAQ

How accurate is Tencent: Hy3 preview?

Tencent: Hy3 preview scores 25.0% overall on the meo benchmark (rank #21 of 22), an un-leaked private holdout graded objective-first.

What is Tencent: Hy3 preview's Effective Value?

Its Effective Value (𝕍) is 0.0000 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 Tencent: Hy3 preview cost per correct answer?

About $0.0140 per correct answer (57,712 tokens, 579 seconds each) at benchmark-time pricing via OpenRouter.

Is Tencent: Hy3 preview open weights?

Yes β€” Tencent: Hy3 preview is released with open weights.