The State of Translation Automation 2026 · Languages

English to Frisian AI translation quality, cost and speed

What it takes to automate translation from English into Frisian against 10 requirements in three levels: the solutions that meet them, what they cost and how fast they are, from the 2026 benchmark of 57 AI models and MT engines.

How good is AI translation from English to Frisian?

Key findings on English to Frisian translation

  1. 1ConstraintsThe two-step solution (translation + post-edit) on OpenAI GPT-5.5 comes nearest to general translation quality with fewer than three steps. It has no critical or major error, but 24.7 weighted errors per 1,000 words, above the bar.See the data →
  2. 2Level step-upThe three-step solution (translation + LQA + repair) on OpenAI GPT-5.5 has the fewest total errors for the Basic requirements, 10.6 per 1,000 words. Its translations have 97.9% of the sentences in the tone of voice, and meeting the Standard requirements takes 100%.See the data →
  3. 3Cost driverEvery level takes three steps (translation + LQA + repair) in Frisian, and the steps and the gender-neutrality agent drive the cost. The solution on Google Gemini 3.1 Pro Preview costs $4.81 per million characters with one step, $560 with three, and $964 with three and the agent for the Advanced requirements.See the data →

Find English to Frisian translation solutions for your requirements

Pick a quality level, or your own requirements, bars and constraints, to see the shortlisted solutions that meet them in Frisian, with a column for each requirement you select. Under Optimize for, choose which of them comes first: the one with the fewest total errors, the lowest cost or the shortest time. Click a column heading to sort by it.

The shortlisted solutions on 13 models, with their Intento LQA results. Total errors per 1,000 words add the weighted general translation quality errors (minor 1, major 5, critical 10) to the missed glossary terms, sentences in the wrong register, tag errors and sentences that fail gender neutrality, each counted once. Formatting counts as met or missed and does not add to the total errors. The models that did not make the shortlist have automatic scores only: the shortlist. Inconclusive: the three LQA judges agree on fewer than 90% of the sentences of the solution, so it has no general translation quality result (09-02). Gender neutrality: the Advanced requirements count the two best solutions of the language, with the gender-neutrality agent where they need it. The results of the other solutions are shown for information.

How we measure it: 11-01 · 11-02

Trade-offs in English to Frisian translation

Trade-off

For the Basic requirements, Optimize for selects the three-step solution (translation + LQA + repair) on OpenAI GPT-5.5 for quality, with 10.6 total errors per 1,000 words. For cost and latency it selects the one on Google Gemini 3.1 Pro Preview. It has 16.8 total errors per 1,000 words, but costs $560 instead of $581 per million characters and takes 281 instead of 320 seconds per 1,000 words.

Constraints

None of the one-step or two-step solutions meet general translation quality in Frisian, and every level includes it. So meeting any level takes three steps. Every solution that meets the Basic requirements costs $560 per million characters or more and takes 280 seconds per 1,000 words or more. None of the three real-time solutions (<2 sec) on the shortlist have a conclusive Intento LQA result.

Agent effect

For the Advanced requirements, the gender-neutrality agent raises the cost of the three-step solution on Google Gemini 3.1 Pro Preview from $560 to $964 per million characters. The time rises from 281 to 301 seconds per 1,000 words. Weighted errors fall from 16.5 to 14.2 per 1,000 words. With the agent the solution has no major error, against two without it. On OpenAI GPT-5.5, the solution with the agent does not meet every Advanced requirement. It has 12.2 total errors per 1,000 words, against 10.6 without it.

English to Frisian translation quality levels

Optimize for
Basic requirements☆☆

Tier 2: two or more solutions, none simple

Solutions that meet them
2 (0 one-step)
Simplest solution
three-step
Fewest errors

OpenAI GPT-5.5
three-step

Total errors per 1,000 words
10.6
Cost per million characters
$581
Seconds per 1,000 words
320
Lowest cost

Google Gemini 3.1 Pro Preview
three-step

Total errors per 1,000 words
16.8
Cost per million characters
$560
Seconds per 1,000 words
281
Shortest time

Google Gemini 3.1 Pro Preview
three-step

Total errors per 1,000 words
16.8
Cost per million characters
$560
Seconds per 1,000 words
281
Standard requirements☆

Tier 3: one solution

Solutions that meet them
1 (0 one-step)
Simplest solution
three-step
Fewest errors

Google Gemini 3.1 Pro Preview
three-step

Total errors per 1,000 words
16.8
Cost per million characters
$560
Seconds per 1,000 words
281
Lowest cost

Google Gemini 3.1 Pro Preview
three-step

Total errors per 1,000 words
16.8
Cost per million characters
$560
Seconds per 1,000 words
281
Shortest time

Google Gemini 3.1 Pro Preview
three-step

Total errors per 1,000 words
16.8
Cost per million characters
$560
Seconds per 1,000 words
281
Advanced requirements☆

Tier 3: one solution

Solutions that meet them
1 (0 one-step)
Simplest solution
with the gender-neutrality agent
Fewest errors

Google Gemini 3.1 Pro Preview
three-step with the gender-neutrality agent

Total errors per 1,000 words
14.2
Cost per million characters
$964
Seconds per 1,000 words
301
Lowest cost

Google Gemini 3.1 Pro Preview
three-step with the gender-neutrality agent

Total errors per 1,000 words
14.2
Cost per million characters
$964
Seconds per 1,000 words
301
Shortest time

Google Gemini 3.1 Pro Preview
three-step with the gender-neutrality agent

Total errors per 1,000 words
14.2
Cost per million characters
$964
Seconds per 1,000 words
301

Optimize for selects one of the solutions that meet the level: the one with the fewest total errors, the lowest cost or the shortest time. Where no solution meets the level, the card shows the available solution with the fewest total errors and what it misses. Total errors count every requirement with countable errors.

Quality levelTierSolutionsSimplest solutionModelTotal errors per 1,000 wordsCost per million charactersSeconds per 1,000 words
BasicTier 2: two or more solutions, none simple☆☆three-stepGoogle Gemini 3.1 Pro Preview16.8$560281
StandardTier 3: one solution☆three-stepGoogle Gemini 3.1 Pro Preview16.8$560281
AdvancedTier 3: one solution☆with the gender-neutrality agentGoogle Gemini 3.1 Pro Preview14.2$964301

One solution per level: the simplest that meets the level, the cheapest of them where several are equally simple. ★ a one-step solution, ☆ a solution with more steps or the gender-neutrality agent.

Level step-up

Every level takes three steps (translation + LQA + repair) in Frisian. Here the Standard requirements add tone of voice, tag handling and formatting to general translation quality. Of the two solutions that meet the Basic requirements, only the one on Google Gemini 3.1 Pro Preview meets these too, with 100% of the sentences in the tone of voice and 100% of the tags in place. The Advanced requirements add gender neutrality. With the gender-neutrality agent, its translations have 100% of the sentences about people gender-neutral.

How we measure it: 11-01 · 11-02 · 08-09

English to Frisian translation cost against errors

Chart data
Cost and weighted errors of the one-step solutions in Frisian
ModelUS dollars per million charactersWeighted errors per 1,000 wordsGroup
OpenAI GPT-5.6 Luna0.4435.9Misses a Standard requirement
DeepSeek V3.20.9380.8Misses a Standard requirement
xAI Grok 4.20 (non-reasoning)1.1490.3Misses a Standard requirement
MoonshotAI Kimi K2.51.25141.7Misses a Standard requirement
z.ai GLM-51.9680.8Misses a Standard requirement
Google Gemini 3.5 Flash3.5948.1Misses a Standard requirement
Google Gemini 3.1 Pro Preview4.8129Misses a Standard requirement
OpenAI GPT-5.510.9628.3Misses a Standard requirement
Anthropic Claude Opus 4.814.7361.3Misses a Standard requirement
Price vs quality

The cheapest one-step solution, on OpenAI GPT-5.6 Luna, costs $0.44 per million characters and has 35.9 weighted errors per 1,000 words. Only two have fewer: the solutions on Google Gemini 3.1 Pro Preview, with 29.0 weighted errors per 1,000 words at $4.81 per million characters, and on OpenAI GPT-5.5, with 28.3 at $10.96. The dearest with a conclusive Intento LQA result, on Anthropic Claude Opus 4.8, costs $14.73 per million characters and has 61.3 weighted errors per 1,000 words.

How we measure it: 11-01 · 10-10

English to Frisian translation requirements

Chart data
Errors per 1,000 words of the one-step solutions in Frisian, median, by requirement
RequirementMedian errors per 1,000 words
General translation quality61.3
Tone of voice0.3
Gender neutrality0.3
Tag handling0

The requirements with countable errors. Every solution with its result on each requirement: the explorer.

Bottleneck

Every one-step solution with a conclusive Intento LQA result is over the bar for weighted errors. Only the ones on OpenAI GPT-5.5 and Google Gemini 3.1 Pro Preview meet the bars for critical and major errors, with 28.3 and 29.0 weighted errors per 1,000 words. With three steps (translation + LQA + repair), they come to 9.6 and 16.5 weighted errors per 1,000 words, under the bar. The three-step solutions on Anthropic Claude Opus 4.8, DeepSeek V3.2 and z.ai GLM-5 have 25.0 to 38.9 weighted errors per 1,000 words.

How we measure it: 09-09 · 09-10 · 09-11 · 09-12 · 09-18

AI models and MT engines for English to Frisian translation

Chart data
Weighted errors per 1,000 words of the shortlisted one-step solutions in Frisian
ModelWeighted errors per 1,000 words of the shortlisted one-step solutions in FrisianClass
MoonshotAI Kimi K2.5141.7LLMs
xAI Grok 4.20 (non-reasoning)90.3LLMs
z.ai GLM-580.8LLMs
DeepSeek V3.280.8LLMs
Anthropic Claude Opus 4.861.3LLMs
Google Gemini 3.5 Flash48.1LLMs
OpenAI GPT-5.6 Luna35.9LLMs
Google Gemini 3.1 Pro Preview29LLMs
OpenAI GPT-5.528.3LLMs
Model types

The nine one-step solutions with a conclusive Intento LQA result in Frisian are all on LLMs. Four of these LLMs meet our language-support criteria: OpenAI GPT-5.5, Google Gemini 3.1 Pro Preview, OpenAI GPT-5.6 Luna and Google Gemini 3.5 Flash. Their translations have 28.3 to 48.1 weighted errors per 1,000 words. Translations from Anthropic Claude Opus 4.8 have 61.3 weighted errors per 1,000 words, just over the limit of 60. The solutions on the MT engine, the LLM-based MT engine and the open-weight LLM have no conclusive result.

How we measure it: 00-04 · 08-15 · 23-01

How we shortlisted English to Frisian translation solutions

Chart data
Ranks of the models on the two automatic metrics, Frisian
ModelCOMET-KiwiRank on COMET-KiwiMetricXRank on MetricXShortlistedShortlist place
Anthropic Claude Fable 50.681102.2811NoNot shortlisted
OpenAI GPT-4.10.679112.229NoNot shortlisted
Anthropic Claude Sonnet 50.673142.3913NoNot shortlisted
Anthropic Claude Haiku 4.50.671152.7918NoNot shortlisted
Anthropic Claude Sonnet 4.60.668162.4014NoNot shortlisted
Mistral Large 30.661172.5916NoNot shortlisted
Google Gemma 40.653193.6723NoNot shortlisted
Swiss AI Apertus-8B-25090.642213.3520NoNot shortlisted
Mistral Medium 3.10.642223.9724NoNot shortlisted
MiniMax M2.50.640233.4621NoNot shortlisted
AI21 Jamba Large 1.70.638243.6622NoNot shortlisted
Alibaba Machine Translation API0.3762510.6925NoNot shortlisted
Cohere Command A Translate0.72612.5415YesTop 8 by COMET-Kiwi
OpenAI GPT-5.6 Luna0.70221.911YesTop 8 by both metrics
OpenAI GPT-5.50.69631.942YesTop 8 by both metrics
Google Gemini 3.5 Flash0.69442.044YesTop 8 by both metrics
Google Translation LLM0.68852.3212YesTop 8 by COMET-Kiwi
z.ai GLM-50.68662.116YesTop 8 by both metrics
xAI Grok 4.20 (non-reasoning)0.68572.095YesTop 8 by both metrics
MoonshotAI Kimi K2.50.68482.2710YesTop 8 by COMET-Kiwi
Anthropic Claude Opus 4.80.68392.167YesTop 8 by MetricX
Google Gemini 3.1 Pro Preview0.678121.963YesTop 8 by MetricX
DeepSeek V3.20.675132.178YesTop 8 by MetricX
Google Cloud Advanced Translation API0.657182.9119YesReal-time place
OpenAI gpt-oss-120b0.648202.6617YesReal-time place
ModelSolution complexityCOMET-Kiwi rankCOMET-KiwiMetricX rankMetricXShortlist
Cohere Command A TranslateOff-the-shelf LLM10.726152.54Top 8 by COMET-Kiwi
OpenAI GPT-5.6 LunaAdjusted LLM20.70211.91Top 8 by both metrics
OpenAI GPT-5.5Adjusted LLM30.69621.94Top 8 by both metrics
Google Gemini 3.5 FlashAdjusted LLM40.69442.04Top 8 by both metrics
Google Translation LLMAdjusted NMT50.688122.32Top 8 by COMET-Kiwi
z.ai GLM-5Adjusted LLM60.68662.11Top 8 by both metrics
xAI Grok 4.20 (non-reasoning)Adjusted LLM70.68552.09Top 8 by both metrics
MoonshotAI Kimi K2.5Adjusted LLM80.684102.27Top 8 by COMET-Kiwi
Anthropic Claude Opus 4.8Adjusted LLM90.68372.16Top 8 by MetricX
Anthropic Claude Fable 5Adjusted LLM100.681112.28Not shortlisted
OpenAI GPT-4.1Adjusted LLM110.67992.22Not shortlisted
Google Gemini 3.1 Pro PreviewAdjusted LLM120.67831.96Top 8 by MetricX
DeepSeek V3.2Adjusted LLM130.67582.17Top 8 by MetricX
Anthropic Claude Sonnet 5Adjusted LLM140.673132.39Not shortlisted
Anthropic Claude Haiku 4.5Adjusted LLM150.671182.79Not shortlisted
Anthropic Claude Sonnet 4.6Adjusted LLM160.668142.40Not shortlisted
Mistral Large 3Adjusted LLM170.661162.59Not shortlisted
Google Cloud Advanced Translation APIOff-the-shelf NMT180.657192.91Real-time place
Google Gemma 4Adjusted LLM190.653233.67Not shortlisted
OpenAI gpt-oss-120bAdjusted LLM200.648172.66Real-time place
Swiss AI Apertus-8B-2509Off-the-shelf LLM210.642203.35Not shortlisted
Mistral Medium 3.1Adjusted LLM220.642243.97Not shortlisted
MiniMax M2.5Adjusted LLM230.640213.46Not shortlisted
AI21 Jamba Large 1.7Adjusted LLM240.638223.66Not shortlisted
Alibaba Machine Translation APIOff-the-shelf NMT250.3762510.69Not shortlisted
OpenAI GPT-5.6 SolAdjusted LLM—0.687—1.95Same model family as OpenAI GPT-5.6 Luna
OpenAI GPT-5.6 TerraAdjusted LLM—0.701—1.82Same model family as OpenAI GPT-5.6 Luna
xAI Grok 4.20 (reasoning)Adjusted LLM—0.681—2.31Same model family as xAI Grok 4.20 (non-reasoning)

Every model we got a translation from, as the shortlist compared it: off the shelf, without context in the request, or adjusted, with it. A model that did not make the shortlist is shown adjusted, unless off the shelf it scores higher by 0.03 or more on COMET-Kiwi. Ranks among 25 models, with each model family counted once, and the top 8 of each metric in bold. Higher is better on COMET-Kiwi, lower on MetricX. The scores rank models within this language and are not a requirement result.

Metrics vs LQA

Google Gemini 3.1 Pro Preview, the model of the only solution that meets every Standard requirement, is on the shortlist through MetricX, 3rd there and 12th on COMET-Kiwi. Cohere Command A Translate, first on COMET-Kiwi and 15th on MetricX, has no conclusive result. Translations from MoonshotAI Kimi K2.5, 8th on COMET-Kiwi, have 141.7 weighted errors per 1,000 words, the most of the solutions with a conclusive result. Nearest the cut in the tail are OpenAI GPT-4.1, 9th on MetricX, and Anthropic Claude Fable 5, 10th on COMET-Kiwi.

How we measure it: 00-08 · 00-09

English to Frisian translation cost

Quality levelCheapest solution that meets every requirement of the levelCost per million characters
BasicGoogle Gemini 3.1 Pro Preview, three-step$560
StandardGoogle Gemini 3.1 Pro Preview, three-step$560
AdvancedGoogle Gemini 3.1 Pro Preview, three-step with the gender-neutrality agent$964
Chart data
Cost of the shortlisted one-step solutions in Frisian
ModelUS dollars per million charactersClass
OpenAI GPT-5.6 Luna$0.44LLMs
DeepSeek V3.2$0.93LLMs
xAI Grok 4.20 (non-reasoning)$1.14LLMs
MoonshotAI Kimi K2.5$1.25LLMs
z.ai GLM-5$1.96LLMs
Google Gemini 3.5 Flash$3.59LLMs
Google Gemini 3.1 Pro Preview$4.81LLMs
OpenAI GPT-5.5$10.96LLMs
Anthropic Claude Opus 4.8$14.73LLMs
Google Cloud Advanced Translation API$20MT engines
Google Translation LLM$20.9LLM-based MT engines
Chart data
Tokens per character of the translation into Frisian, by model
ModelTokens per characterClass
Anthropic Claude Fable 50.486LLMs
Anthropic Claude Opus 4.80.485LLMs
Anthropic Claude Sonnet 50.483LLMs
Mistral Medium 3.10.394LLMs
Mistral Large 30.39LLMs
Anthropic Claude Sonnet 4.60.383LLMs
AI21 Jamba Large 1.70.38LLMs
Anthropic Claude Haiku 4.50.378LLMs
MoonshotAI Kimi K2.50.349LLMs
DeepSeek V3.20.347LLMs
z.ai GLM-50.344LLMs
Google Gemini 3.1 Pro Preview0.333LLMs
Google Gemini 3.5 Flash0.33LLMs
Google Gemma 40.326Open-weight LLMs
xAI Grok 4.20 (reasoning)0.314LLMs
xAI Grok 4.20 (non-reasoning)0.313LLMs
Cohere Command A Translate0.311LLMs
OpenAI GPT-5.50.305LLMs
OpenAI GPT-5.6 Sol0.304LLMs
OpenAI GPT-5.6 Terra0.304LLMs
OpenAI GPT-4.10.303LLMs
OpenAI GPT-5.6 Luna0.303LLMs
MiniMax M2.50.302LLMs

How we measure it: 10-01 · 10-03 · 05-02

English to Frisian translation speed

Quality levelFastest solution that meets every requirement of the levelSeconds per 1,000 words
BasicGoogle Gemini 3.1 Pro Preview, three-step281
StandardGoogle Gemini 3.1 Pro Preview, three-step281
AdvancedGoogle Gemini 3.1 Pro Preview, three-step with the gender-neutrality agent301
Chart data
Time to translate 1,000 words into Frisian, by model
ModelSeconds per 1,000 wordsClass
Google Cloud Advanced Translation API0.2MT engines
Google Translation LLM1.4LLM-based MT engines
OpenAI gpt-oss-120b2.3Open-weight LLMs
OpenAI GPT-5.6 Luna17LLMs
Google Gemini 3.5 Flash18LLMs
xAI Grok 4.20 (non-reasoning)21LLMs
Alibaba Machine Translation API26MT engines
OpenAI GPT-5.6 Terra41LLMs
xAI Grok 4.20 (reasoning)43LLMs
Anthropic Claude Haiku 4.545LLMs
Swiss AI Apertus-8B-250946Open-weight LLMs
Cohere Command A Translate48LLMs
Mistral Medium 3.151LLMs
AI21 Jamba Large 1.752LLMs
MiniMax M2.553LLMs
OpenAI GPT-4.156LLMs
Mistral Large 368LLMs
Anthropic Claude Sonnet 574LLMs
Anthropic Claude Fable 581LLMs
Anthropic Claude Sonnet 4.684LLMs
Google Gemini 3.1 Pro Preview88LLMs
DeepSeek V3.288LLMs
Anthropic Claude Opus 4.8100LLMs
OpenAI GPT-5.6 Sol100LLMs
OpenAI GPT-5.5128LLMs
Google Gemma 4149Open-weight LLMs
z.ai GLM-5245LLMs
MoonshotAI Kimi K2.5248LLMs

Real-time models are those the shortlist counts as real-time, under 2 seconds per text across the languages we timed. The chart shows the time of each model into Frisian.

Real-time

The three real-time models in Frisian are an MT engine, Google Cloud Advanced Translation API, an LLM-based MT engine, Google Translation LLM, and an open-weight LLM, OpenAI gpt-oss-120b. Their translations have 94.3%, 95.1% and 99.6% of the tags in place. Next come translations from OpenAI GPT-5.6 Luna, at 17 seconds per 1,000 words. The fastest solution that meets the Basic or Standard requirements takes 281 seconds per 1,000 words, and the one that meets the Advanced requirements 301, with the gender-neutrality agent.

How we measure it: 12-01 · 12-02 · 12-09

Tone of voice and gender-neutral English to Frisian translation

Tone of voice
Formal register, with jo
Gender-neutrality strategy
neutral nouns + rephrasing
Options in the language
generic hy, hy of sy, neutral nouns + rephrasing
Default gender
None where the source leaves the gender open
The two best solutionsSolutionSentences about people that are gender-neutralMeets every Advanced requirementCost per million charactersSeconds per 1,000 words
Google Gemini 3.1 Pro Previewthree-step, with the gender-neutrality agent100%✓$964301
OpenAI GPT-5.5three-step, with the gender-neutrality agent99.1%✗$978313
Analysis

The tone of voice we check in Frisian is the formal register, with "jo" as the form of address. For gender neutrality we check neutral nouns and rephrasing, with no default gender. The other options in the language are generic "hy" and "hy of sy". With the gender-neutrality agent, translations from the three-step solutions (translation + LQA + repair) on Google Gemini 3.1 Pro Preview and OpenAI GPT-5.5 have 100% and 97.9% of the sentences in the tone of voice.

How we measure it: 02-05 · 02-06 · 08-11 · 09-18

How confident are we in measuring English to Frisian translation quality?

Confidence

We shortlist the models on COMET-Kiwi and MetricX, the quality-estimation pair that agrees in Frisian, and the result is not borderline. The three LQA judges agree on a median of 92.1% of the sentences across the 13 shortlisted solutions. So the ranking of the shortlist and the Intento LQA results on this page carry weight, even with the similarity pair short of its threshold.

How we measure it: 00-07 · 07-05 · 09-02

Automate translation to meet your requirements

See Intento Language Hub in action.

Book a demo
Leave feedback