InterMIND
Translation quality, in the open

AI Translation Quality Benchmark

How well InterMIND actually translates — scored automatically from live demo traffic, per language pair, every month. No cherry-picked averages.

Chat translation
98/100
+1 vs prev median · 13 pairs · 39 runs · May 2026

typed messages, translated inline

Voice translation
68/100
-5 vs prev median · 26 pairs · 78 runs · May 2026

one score for speech recognition + translation, end-to-end

Try a live translation Read the methodology
39 language pairs·117 scored runs·14 languages·updated monthly · May 2026

Score = semantic fidelity to a reference translation (0–100). Higher is better.

Excellent 90+ Good 70–89 Fair 50–69 Weak <50

Text chat translation

Typed messages translated inline as they're sent.

20 of 21 languages average Good or better (70+) this quarter · 11 improved vs Q1 2026

Average quality per language · Q2 2026 vs Q1 2026

Each bar pools every run translating to or from that language this quarter. The dashed notch marks the previous quarter.

Q2 2026 (avg) Q1 2026 median 98 · May 2026
100
75
50
25
0
100
100
97
94
93
92
90
87
84
84
84
83
82
80
80
79
77
76
76
75
66
—sv
—da
—ro
—fi
▲21ko
—no
▲17en
▲18nl
▲12ru
▲5de
▲11it
•es
—is
▲3fr
▲3pt
▲2pl
▲1hu
•cs
▲1ja
▼2zh
—uk

Strongest pairs · May 2026

en → de
100
100–100 · 3 runs
en → ja
100
98–100 · 3 runs
en → ko
100
100–100 · 3 runs
en → pt
100
100–100 · 3 runs
en → zh
100
99–100 · 3 runs
en → es
99
99–100 · 3 runs

Voice translation

Live speech, transcribed and translated in real time — the harder end-to-end problem.

10 of 21 languages average Good or better (70+) this quarter · 14 improved vs Q1 2026

Average quality per language · Q2 2026 vs Q1 2026

Each bar pools every run translating to or from that language this quarter. The dashed notch marks the previous quarter.

Q2 2026 (avg) Q1 2026 median 68 · May 2026
100
75
50
25
0
82
81
76
76
76
75
74
72
70
70
69
69
68
68
68
67
67
67
64
62
62
▲22ko
—ro
▲9ja
▲18en
—sv
—uk
▲6zh
▲10es
▲22pl
▲20nl
▲9hu
▲2pt
—no
▲16cs
—is
—da
▲14it
▲15de
▲6fr
▲4ru
—fi

Strongest pairs · May 2026

en → zh
91
43–98 · 3 runs
en → pt
81
76–81 · 3 runs
en → pl
79
50–91 · 3 runs
en → es
78
58–82 · 3 runs
en → ko
78
70–99 · 3 runs
ru → ko
78
66–80 · 3 runs

Quality over time

Monthly median per modality, since we started publishing.

Chat Voice
100
75
50
25
0
62
98
73
98
68
Mar 2026
Apr 2026
May 2026

These numbers come from real sessions

Every run on the live demo is scored automatically and folded into next month's benchmark. A session looks like this:

Live translated meeting
→ EN
Try a live translation Get started free