Teanga Bench

Gemini 4 Argon

Google · google/gemini-4-argon · run 16 November 2026 · each item asked 3 times

Overall score
81.9%95% range 80.5–83.3%
Rank
4 of 9
Cost per run
$3.10this suite, judges included
Per 1,000 correct
$0.83cost of 1,000 full-score answers
Consistency
72.5%same score in all 3 asks, mean over tests

Items and answers come from the named sources and have not been reviewed by people, so a few gold answers may be wrong. How the items are built

Scores by test

#

Gemini 4 Argon places best in Lesson exercises (2nd of 9, tied with the top) and worst in Word forms (5th of 9, 8.7 points below the top); tied means the 95% ranges overlap.

Score on each test. Right: the score and its place among the ranked models. Select a test to compare every model.

  • This model, with its 95% range
  • Each other ranked model
  1. Lesson exercisesClaude Fable 5.2: 98.5%Gemini 4 Hydrogen: 97.3%GPT-6 Astra: 96.6%DeepSeek V4.1 Flash 1017: 95.6%Kimi K3.1 Pro: 95.8%GPT-6 Luna: 93.6%Claude Haiku 5.5: 87.5%Qwen3.8 Flash: 90.6%Gemini 4 Argon: 98.2%, 95% range 96.9% to 99.3%, 2nd of 9
  2. Word formsClaude Fable 5.2: 94.1%Gemini 4 Hydrogen: 91.1%GPT-6 Astra: 92.5%DeepSeek V4.1 Flash 1017: 87.0%Kimi K3.1 Pro: 82.1%GPT-6 Luna: 80.5%Claude Haiku 5.5: 65.3%Qwen3.8 Flash: 50.6%Gemini 4 Argon: 85.4%, 95% range 82.1% to 88.5%, 5th of 9
  3. TranslationClaude Fable 5.2: 75.1%Gemini 4 Hydrogen: 72.4%GPT-6 Astra: 73.1%DeepSeek V4.1 Flash 1017: 68.8%Kimi K3.1 Pro: 69.4%GPT-6 Luna: 68.0%Claude Haiku 5.5: 57.9%Qwen3.8 Flash: 56.9%Gemini 4 Argon: 71.9%, 95% range 69.4% to 74.4%, 4th of 9
  4. Round tripClaude Fable 5.2: 74.3%Gemini 4 Hydrogen: 73.0%GPT-6 Astra: 69.3%DeepSeek V4.1 Flash 1017: 68.5%Kimi K3.1 Pro: 65.3%GPT-6 Luna: 64.5%Claude Haiku 5.5: 58.0%Qwen3.8 Flash: 52.2%Gemini 4 Argon: 70.0%, 95% range 67.3% to 72.6%, 3rd of 9
  5. Tutor accuracyClaude Fable 5.2: 88.5%Gemini 4 Hydrogen: 88.1%GPT-6 Astra: 86.5%DeepSeek V4.1 Flash 1017: 76.1%Kimi K3.1 Pro: 79.3%GPT-6 Luna: 71.3%Claude Haiku 5.5: 66.9%Qwen3.8 Flash: 58.0%Gemini 4 Argon: 84.1%, 95% range 78.9% to 88.9%, 4th of 9
View as table
TestScore and 95% rangePlaceHighestMedian model
Lesson exercises98.2% 96.9–99.3%2nd of 9Claude Fable 5.2 (98.5%)95.8%
Word forms85.4% 82.1–88.5%5th of 9Claude Fable 5.2 (94.1%)85.4%
Translation71.9% 69.4–74.4%4th of 9Claude Fable 5.2 (75.1%)69.4%
Round trip70.0% 67.3–72.6%3rd of 9Claude Fable 5.2 (74.3%)68.5%
Tutor accuracy84.1% 78.9–88.9%4th of 9Claude Fable 5.2 (88.5%)79.3%

Gemini 4 Argon places best in Lesson exercises (2nd of 9, tied with the top) and worst in Word forms (5th of 9, 8.7 points below the top); tied means the 95% ranges overlap.

One axis per test, with the same scale on every profile: 0% at the centre to 100% at the outer ring. Solid accent: this model; faint dashed polygon: median model score on each test; faint dotted outer polygon: best score on each test (not necessarily one model). Vertex labels show this model’s scores. Compare scores, not polygon area; no confidence intervals are shown.

Gemini 4 Argon test profileScale: 0% at the centre to 100% at the outer ring. Lesson exercises: 98.2%, median 95.8%, best 98.5%; Word forms: 85.4%, median 85.4%, best 94.1%; Translation: 71.9%, median 69.4%, best 75.1%; Round trip: 70.0%, median 68.5%, best 74.3%; Tutor accuracy: 84.1%, median 79.3%, best 88.5%98.2%Lessonexercises85.4%Word forms71.9%Translation70.0%Round trip84.1%Tutor accuracy
View as table
TestThis modelMedianBest
Lesson exercises98.2%95.8%98.5%
Word forms85.4%85.4%94.1%
Translation71.9%69.4%75.1%
Round trip70.0%68.5%74.3%
Tutor accuracy84.1%79.3%88.5%

Items, consistency and cost by test

Consistency is the share of items that got the same score in every repeat ask.

TestScore95% rangeItemsConsistencyCostPer 1,000 correct
Lesson exercises98.2%96.9–99.3%36098.1%$0.58$0.55
Word forms85.4%82.1–88.5%36089.4%$0.43$0.46
Translation71.9%69.4–74.4%36051.7%$0.44$0.57
Round trip70.0%67.3–72.6%24032.1%$0.59$1.16
Tutor accuracy84.1%78.9–88.9%18091.1%$1.06$2.34

By language

Score, 95% range and number of items (n) for each test.

LanguageLesson exercisesWord formsTranslationRound tripTutor accuracy
Irish95.0% 90.0–98.9% · n=6088.3% 80.0–95.6% · n=6081.5% 76.8–85.6% · n=6082.8% 77.8–87.6% · n=4088.9% 81.7–95.0% · n=60
Welsh99.4% 97.8–100.0% · n=6091.1% 83.9–97.2% · n=6079.8% 75.0–84.2% · n=6082.4% 77.6–86.9% · n=4088.3% 80.6–95.0% · n=60
Scottish Gaelic98.3% 95.0–100.0% · n=6087.8% 80.0–94.4% · n=6076.6% 70.7–82.4% · n=6074.9% 69.7–80.4% · n=4075.0% 63.9–85.0% · n=60
Manx99.4% 98.3–100.0% · n=6082.8% 73.9–90.6% · n=6048.9% 43.6–54.6% · n=6055.3% 50.7–60.2% · n=40n/a
Breton98.3% 95.0–100.0% · n=6087.2% 79.4–93.9% · n=6069.5% 63.9–75.2% · n=6060.7% 54.6–67.2% · n=40n/a
Cornish98.9% 96.7–100.0% · n=6075.0% 64.4–84.4% · n=6075.3% 69.5–81.2% · n=6063.8% 57.5–70.3% · n=40n/a

By CEFR level

The CEFR level comes from the course content; items without one are listed as unspecified.

Score, 95% range and number of items (n) for each test.

CEFR levelLesson exercisesWord formsTranslationRound tripTutor accuracy
A1100.0% 100.0–100.0% · n=12584.8% 79.1–90.0% · n=13474.5% 69.5–79.3% · n=11268.2% 59.4–77.2% · n=2590.9% 80.3–100.0% · n=22
A296.8% 94.3–99.0% · n=13585.5% 80.0–90.5% · n=14068.0% 62.9–73.1% · n=9169.9% 65.5–74.5% · n=8887.7% 75.3–97.5% · n=27
B197.3% 93.2–100.0% · n=7379.5% 67.5–90.6% · n=3965.0% 58.5–71.7% · n=4767.0% 62.6–71.5% · n=7977.0% 64.3–88.1% · n=42
B2100.0% 100.0–100.0% · n=2481.2% 62.5–95.8% · n=1666.9% 57.6–75.9% · n=3069.0% 61.4–76.6% · n=2773.9% 56.5–91.3% · n=23
C1100.0% 100.0–100.0% · n=3n/a83.7% 77.9–89.0% · n=1482.9% 74.9–90.3% · n=1583.3% 60.0–100.0% · n=10
C2n/an/a79.5% 72.5–88.2% · n=689.2% 78.9–99.2% · n=6100.0% 100.0–100.0% · n=2
Unspecifiedn/a96.8% 90.3–100.0% · n=3177.4% 72.6–81.8% · n=60n/a88.9% 81.5–95.7% · n=54

Grammar topics

The topic is the grammar point or error type recorded on each item.

Lesson exercises

Strongest 5 topics.

TopicScoreItems
adjectives100.0%5
chapter overview100.0%11
daily life100.0%5
food100.0%5
grammar point100.0%10

Weakest 5 topics.

TopicScoreItems
vocabulary85.7%7
conversation88.9%6
grammar95.2%21
recap98.2%37
guide99.0%32

Only topics with at least 5 items are ranked.

Word forms

Strongest 5 topics.

TopicScoreItems
verb preterite100.0%5
article gender mutation96.8%31
prepositional pronoun96.4%28
verb imperative92.6%27
number personal92.6%9

Weakest 5 topics.

TopicScoreItems
number counting40.0%5
verb pluperfect66.7%5
verb imperfect73.8%14
verb future75.4%23
noun plural77.8%30

Only topics with at least 5 items are ranked.

Translation

All topics, highest score first.

TopicScoreItems
legal or administrative77.4%60
to english75.3%150
from english66.3%150

Tutor accuracy

All topics, highest score first.

TopicScoreItems
unit focus96.7%10
orthography doubled letter95.2%7
guide91.7%20
grammar90.5%42
no error82.2%30
vocabulary73.3%15
grammar point69.4%12
application54.5%11

Only topics with at least 5 items are ranked.

Tutor explanations

judge-marked, secondary

Tutor items also ask for a short explanation. Two judge models from other model families mark it. This is a secondary number: it is judge-marked and not human-validated.

Tutor accuracy

Headline score
84.1%: found the error, fixed it and changed nothing else (rule-only, no judge).
Parts of the headline
Found 85.0% · fixed 84.1% · no over-correction 89.1%
Explanation score
75.9% (95% range 70.1–81.7%), judge-marked, not human-validated. An explanation passes only when both judges pass it.
Judge agreement
95.2% of 505 judged answers
Unscored
35 answers a judge could not mark, after 37 failed judge calls; they are left out, not scored 0.
Judges
deepseek/deepseek-v4.1-flash-1017, qwen/qwen3.8-flash

Cost and prices

This suite
$3.10, of which $0.29 was judge spend
Whole run
$5.57 for both suites, against a cap of $25.00
Cost per correct answer
$0.000834 ($0.83 per 1,000): suite cost divided by the number of full-score answers
Run date
16 November 2026
Prices for google/gemini-4-argon
$2.00 input, $10.00 output, $0.1000 cache read, $0.19 cache write per million tokens. Retrieved 16 November 2026 from https://openrouter.ai/api/v1/models.
Judge deepseek/deepseek-v4.1-flash-1017
$0.30 input, $1.20 output, $0.0060 cache read per million tokens. Retrieved 16 November 2026 from OpenRouter.
Judge qwen/qwen3.8-flash
$0.15 input, $0.47 output, $0.0160 cache read per million tokens. Retrieved 16 November 2026 from OpenRouter.

Failure examples

11

The lowest-scoring items from the public samples, with the worst of the repeats. Private items are never shown. Model output is shown exactly as text.

Word forms

blas-morph:manx:number_ordinal:gv-num-12

Manx · number ordinal · score 0.00

Prompt

Language: Manx
Task: Give the ordinal numeral for the number below: the ordinal ('first', 'second', ...), including the definite article yn as it stands before the ordinal.
Number: 12 (twelve)
Reply with the answer on one line, in this exact form:
ANSWER: <answer>

Expected

yn nah jeig

Model output

ANSWER: yn ghaa-yeigoo

blas-morph:scots-gaelic:copula:cop-gd-empha-04

Scottish Gaelic · copula · score 0.00

Prompt

Language: Scottish Gaelic
Task: Fill in the blank with the copula word (or copula particle) that the sentence needs. Write only the missing word(s).
Sentence: ___ do Dhia a bhuineas slàinte.
English: Salvation is God's affair.
Reply with the answer on one line, in this exact form:
ANSWER: <answer>

Expected

'S ann

Model output

ANSWER: Is ann

blas-morph:scots-gaelic:number_counting:gd-num-2

Scottish Gaelic · number counting · score 0.00

Prompt

Language: Scottish Gaelic
Task: Give the counting numeral for the number below: the form used when counting aloud (not before a noun).
Number: 2 (two)
Reply with the answer on one line, in this exact form:
ANSWER: <answer>

Expected

dhà

Model output

ANSWER: a dhà

Translation

blas-translation-scots-gaelic-gd-sent-0223-from_english

Scottish Gaelic · from english · score 0.15

Prompt

Translate this English sentence into Scottish Gaelic.
Reply with the Scottish Gaelic translation only, on one line, with no notes or explanation.

English: He is a good carpenter.

Expected

Tha e na dheagh shaor.

Model output

Is e saor math a th' ann.

blas-translation-manx-gv-sent-parallel-15649-from_english

Manx · from english · score 0.30

Prompt

Translate this English sentence into Manx.
Reply with the Manx translation only, on one line, with no notes or explanation.

English: Have you made my bed?

Expected

Vel shiu er chiartaghey my lhiabbee?

Model output

Vel y lhiabbee aym jeant ayd?

blas-translation-cornish-kw-sent-tat-1356-from_english

Cornish · from english · score 0.25

Prompt

Translate this English sentence into Cornish.
Reply with the Cornish translation only, on one line, with no notes or explanation.

English: Do you like dogs?

Expected

Yw da genes keun?

Model output

A garowgh hwi keun?

Round trip

blas-round_trip-cornish-kw-sent-tat-1830

Cornish · round trip · score 0.21

Prompt

Translate this Cornish sentence into English.
Reply with the English translation only, on one line, with no notes or explanation.

Cornish: Kas yw genev an traow na.

Expected

Kas yw genev an traow na.

Model output

Yma kas dhymm a'n taklow na.

Pivot text

I hate those things.

blas-round_trip-manx-gv-sent-parallel-6412

Manx · round trip · score 0.21

Prompt

Translate this Manx sentence into English.
Reply with the English translation only, on one line, with no notes or explanation.

Manx: Nee eh costal queig keead stiagh as magh.

Expected

Nee eh costal queig keead stiagh as magh.

Model output

Bee eh kuse queig cheead ny smoo ny ny sloo.

Pivot text

It will cost five hundred more or less.

blas-round_trip-manx-gv-sent-parallel-2423

Manx · round trip · score 0.34

Prompt

Translate this Manx sentence into English.
Reply with the English translation only, on one line, with no notes or explanation.

Manx: Dreill eh smaght er eh hene.

Expected

Dreill eh smaght er eh hene.

Model output

Chum eh smacht er hene.

Pivot text

He controlled himself.

Tutor accuracy

blas-tutor-irish-clean-sent-9f6457b394

Irish · no error · score 0.00

Prompt

You are a Irish language tutor checking a sentence written by a learner.
Language: Irish
The sentence may contain one error, or it may already be correct.
If it has an error, change only that one error: give the exact erroneous substring and the replacement for it. If it is correct, set has_error to false, leave error_span and correction empty and repeat the sentence as corrected_sentence.
Sentence: Tá sé tábhachtach nach sárófaí a gcearta daonna.
Reply with one JSON object and nothing else:
{"…

Expected

Tá sé tábhachtach nach sárófaí a gcearta daonna.

Model output

{"has_error": true, "error_span": "sárófaí", "correction": "sárófar", "corrected_sentence": "Tá sé tábhachtach nach sárófar a gcearta daonna.", "explanation": "Because the main verb 'Tá' is in the present tense, the verb in the subordinate clause should be in the future tense ('sárófar') rather than the conditional mood ('sárófaí') to maintain the correct sequence of tenses."}

blas-tutor-welsh-exd-bc34d0d6aa

Welsh · unit focus · score 0.00

Prompt

You are a Welsh language tutor checking a sentence written by a learner.
Language: Welsh
The sentence may contain one error, or it may already be correct.
If it has an error, change only that one error: give the exact erroneous substring and the replacement for it. If it is correct, set has_error to false, leave error_span and correction empty and repeat the sentence as corrected_sentence.
Sentence: Mae'r dŵr yn glân.
Reply with one JSON object and nothing else:
{"has_error": true|false, "error…

Expected

Mae'r dŵr yn lân.

Model output

{"has_error": true, "error_span": "yn glân", "correction": "yn lân", "corrected_sentence": "Mae'r dŵr yn lân.", "explanation": "Adjectives undergo a soft mutation after the predicative particle 'yn', so 'glân' becomes 'lân'."}