The State of Translation Automation 2026 · Languages
What it takes to automate translation from English into Irish 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.
A solution is a model plus its harness: prompts, glossaries and other context, pre- and post-processing. It meets a quality level when it meets every requirement of the level.
| Requirement | Description | Met when |
|---|---|---|
| Basic requirements for any company | ||
| General translation quality | Translation errors of every type, such as mistranslations, omissions and grammar errors, classified by type and severity following MQM, the industry standard. | No critical error, at most 3 major errors in the four test texts (3,034 words) and at most 20 errors per 1,000 words, weighted by severity (minor 1, major 5, critical 10). |
| Standard requirements adds the rules companies set for their content | ||
| Terminology | The approved translation of every glossary term, in any of its correct forms. | 100% of the glossary terms. We test it in the 16 languages with a glossary, so it does not apply in Irish. |
| Tone of voice | The formal register in every sentence in the second person or the imperative. | 100% of those sentences. |
| Tags | Every HTML tag of the source, with its attributes, in the translation. | 100% of the tags. |
| Unicode character integrity | Every character intact, with no corrupted characters and no letter from another script that only looks like the right one. | 100% of the characters. |
| Locale conventions | The punctuation marks of the target language in place of the English ones. | No pass-or-fail bar. We report it as a share, because whether to localize punctuation is a matter of house style. |
| HTML entities | No HTML character code, such as the code for a quotation mark, that the source does not carry. | No added HTML entity. |
| Valid JSON | The translated JSON document still parses. | The document parses. |
| Spacing at tag boundaries | The spaces next to HTML tags stay as in the source. | 100% of the tag boundaries. |
| Advanced requirements adds a requirement for some companies and use cases | ||
| Gender neutrality | A gender-neutral translation wherever the English source leaves the gender of a person open, as in “your manager” or “the expert”. | 100% of the sentences that refer to a person. |
Pick a quality level, or your own requirements, bars and constraints, to see the shortlisted solutions that meet them in Irish, 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 12 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.
Five three-step solutions (translation + LQA + repair) meet every Standard requirement in Irish. None of them leads on errors, cost and time at once. The one on Google Gemini 3.1 Pro Preview has the fewest total errors, 7.2 per 1,000 words, and takes 324 seconds per 1,000 words. The one on OpenAI GPT-5.6 Luna is the cheapest and the fastest, at 139 seconds per 1,000 words, with 14.5 total errors per 1,000 words. Optimizing for quality selects Google Gemini 3.1 Pro Preview, and optimizing for cost or latency OpenAI GPT-5.6 Luna.
Of the five solutions that meet every Standard requirement in Irish, a cost cap from $562 to $583 per million characters leaves two, those on OpenAI GPT-5.6 Luna and Google Gemini 3.1 Pro Preview. Only the one on OpenAI GPT-5.6 Luna, at 139 seconds per 1,000 words, fits a time limit under 290. The solutions on Anthropic Claude Fable 5, Anthropic Claude Opus 4.8 and OpenAI GPT-5.5 have 15.2 to 16.5 total errors per 1,000 words. They have more errors, cost more and take longer than the one on OpenAI GPT-5.6 Luna.
For the Advanced requirements, the gender-neutrality agent works with the three-step solutions (translation + LQA + repair) on Google Gemini 3.1 Pro Preview and OpenAI GPT-5.6 Luna. On Google Gemini 3.1 Pro Preview it raises the cost from $561 to $951 per million characters, and total errors rise from 7.2 to 8.6 per 1,000 words. On OpenAI GPT-5.6 Luna the cost rises from $548 to $963 per million characters, total errors from 14.5 to 17.5 per 1,000 words and the time from 139 to 141 seconds per 1,000 words.
Google Gemini 3.1 Pro Preview
three-step
OpenAI GPT-5.6 Luna
three-step
OpenAI GPT-5.6 Luna
three-step
Google Gemini 3.1 Pro Preview
three-step
OpenAI GPT-5.6 Luna
three-step
OpenAI GPT-5.6 Luna
three-step
Google Gemini 3.1 Pro Preview
three-step with the gender-neutrality agent
Google Gemini 3.1 Pro Preview
three-step with the gender-neutrality agent
OpenAI GPT-5.6 Luna
three-step with the gender-neutrality agent
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 level | Tier | Solutions | Simplest solution | Model | Total errors per 1,000 words | Cost per million characters | Seconds per 1,000 words |
|---|---|---|---|---|---|---|---|
| Basic | Tier 2: two or more solutions, none simple | ☆×5 | three-step | OpenAI GPT-5.6 Luna | 14.5 | $548 | 138 |
| Standard | Tier 2: two or more solutions, none simple | ☆×5 | three-step | OpenAI GPT-5.6 Luna | 14.5 | $548 | 138 |
| Advanced | Tier 2: two or more solutions, none simple | ☆☆ | with the gender-neutrality agent | Google Gemini 3.1 Pro Preview | 8.6 | $951 | 320 |
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.
The same five three-step solutions (translation + LQA + repair) meet every Basic and every Standard requirement in Irish. The Advanced requirements keep two, on Google Gemini 3.1 Pro Preview and OpenAI GPT-5.6 Luna with the gender-neutrality agent, the two best, the only ones the Advanced requirements count. Optimize for selects Google Gemini 3.1 Pro Preview for quality and OpenAI GPT-5.6 Luna for latency under all three. For cost it selects OpenAI GPT-5.6 Luna, and Google Gemini 3.1 Pro Preview under the Advanced requirements.
| Model | US dollars per million characters | Weighted errors per 1,000 words | Group |
|---|---|---|---|
| OpenAI GPT-5.6 Luna | 0.49 | 43.2 | Misses a Standard requirement |
| xAI Grok 4.20 (reasoning) | 1.34 | 67.9 | Misses a Standard requirement |
| Google Gemini 3.5 Flash | 4.12 | 46.5 | Misses a Standard requirement |
| Google Gemini 3.1 Pro Preview | 5.5 | 25 | Misses a Standard requirement |
| OpenAI GPT-5.5 | 12.19 | 31 | Misses a Standard requirement |
| Anthropic Claude Opus 4.8 | 16.69 | 55.4 | Misses a Standard requirement |
| Google Translation LLM | 21.94 | 84.7 | Misses a Standard requirement |
| Translated Lara 2 | 24.99 | 80.1 | Misses a Standard requirement |
| DeepL Next Gen | 27.5 | 91.6 | Misses a Standard requirement |
| Anthropic Claude Fable 5 | 33.43 | 43.8 | Misses a Standard requirement |
The front has two three-step solutions (translation + LQA + repair) on LLMs, OpenAI GPT-5.6 Luna and Google Gemini 3.1 Pro Preview. The other three that meet every Standard requirement cost more, $584 to $617 per million characters, and have more weighted errors, 15.2 to 15.8 per 1,000 words. One-step solutions on the three shortlisted LLM-based MT engines cost $21.94 to $27.50 per million characters and have 80.1 to 91.6 weighted errors per 1,000 words, against $5.50 and 25.0 on Google Gemini 3.1 Pro Preview.
| Requirement | Median errors per 1,000 words |
|---|---|
| General translation quality | 51 |
| Gender neutrality | 0.3 |
| Tone of voice | 0 |
| Tag handling | 0 |
The requirements with countable errors. Every solution with its result on each requirement: the explorer.
General translation quality keeps every one-step and two-step solution with a conclusive result out. The solutions on Google Gemini 3.1 Pro Preview come closest to the bar, at 25.0 weighted errors per 1,000 words in one step and 24.7 in two (translation + post-edit). Three-step solutions (translation + LQA + repair) on five of the seven models have 6.9 to 15.8 weighted errors per 1,000 words, under the bar. Translations from the three shortlisted LLM-based MT engines have 94.5% to 99.8% of tags right, where tag handling requires 100%.
| Model | Weighted errors per 1,000 words of the shortlisted one-step solutions in Irish | Class |
|---|---|---|
| DeepL Next Gen | 91.6 | LLM-based MT engines |
| Google Translation LLM | 84.7 | LLM-based MT engines |
| Translated Lara 2 | 80.1 | LLM-based MT engines |
| xAI Grok 4.20 (reasoning) | 67.9 | LLMs |
| Anthropic Claude Opus 4.8 | 55.4 | LLMs |
| Google Gemini 3.5 Flash | 46.5 | LLMs |
| Anthropic Claude Fable 5 | 43.8 | LLMs |
| OpenAI GPT-5.6 Luna | 43.2 | LLMs |
| OpenAI GPT-5.5 | 31 | LLMs |
| Google Gemini 3.1 Pro Preview | 25 | LLMs |
None of the one-step translations from the seven shortlisted LLMs with a conclusive Intento LQA result in Irish have a critical error. They have 25.0 to 67.9 weighted errors per 1,000 words. Translations from the three shortlisted LLM-based MT engines have 80.1 to 91.6 weighted errors per 1,000 words and 1 to 8 critical errors each. Of the seven LLMs, xAI Grok 4.20 (reasoning) is the one outside our language-support criteria, at 67.9 weighted errors per 1,000 words.
| Model | COMET-Kiwi | Rank on COMET-Kiwi | MetricX | Rank on MetricX | Shortlisted | Shortlist place |
|---|---|---|---|---|---|---|
| DeepSeek V3.2 | 0.709 | 12 | 2.17 | 13 | No | Not shortlisted |
| Anthropic Claude Sonnet 5 | 0.709 | 13 | 2.16 | 12 | No | Not shortlisted |
| MoonshotAI Kimi K2.5 | 0.706 | 14 | 2.48 | 19 | No | Not shortlisted |
| Mistral Large 3 | 0.706 | 15 | 2.23 | 14 | No | Not shortlisted |
| Anthropic Claude Sonnet 4.6 | 0.705 | 16 | 2.13 | 11 | No | Not shortlisted |
| Tilde TildeOpen LLM engine | 0.701 | 18 | 2.32 | 16 | No | Not shortlisted |
| Microsoft Translator | 0.698 | 19 | 2.34 | 17 | No | Not shortlisted |
| Google Cloud Advanced Translation API | 0.689 | 20 | 3.01 | 25 | No | Not shortlisted |
| OpenAI gpt-oss-120b | 0.689 | 21 | 2.79 | 23 | No | Not shortlisted |
| Mistral Medium 3.1 | 0.684 | 22 | 2.67 | 20 | No | Not shortlisted |
| ModernMT Enterprise Edition | 0.681 | 23 | 2.97 | 24 | No | Not shortlisted |
| Yandex Cloud Translate v2 | 0.681 | 24 | 2.77 | 21 | No | Not shortlisted |
| Anthropic Claude Haiku 4.5 | 0.679 | 25 | 2.78 | 22 | No | Not shortlisted |
| Google Gemma 4 | 0.670 | 26 | 4.49 | 30 | No | Not shortlisted |
| Swiss AI Apertus-8B-2509 | 0.658 | 27 | 3.09 | 26 | No | Not shortlisted |
| MiniMax M2.5 | 0.649 | 28 | 3.60 | 28 | No | Not shortlisted |
| Amazon Translate | 0.644 | 29 | 3.93 | 29 | No | Not shortlisted |
| Cohere Tiny Aya | 0.640 | 30 | 3.57 | 27 | No | Not shortlisted |
| Alibaba Qwen3-235B-A22B | 0.581 | 31 | 5.36 | 31 | No | Not shortlisted |
| Cohere Command A Translate | 0.520 | 32 | 8.01 | 32 | No | Not shortlisted |
| AI21 Jamba Large 1.7 | 0.474 | 33 | 9.59 | 33 | No | Not shortlisted |
| xAI Grok 4.20 (reasoning) | 0.724 | 1 | 1.96 | 4 | Yes | Top 8 by both metrics |
| OpenAI GPT-5.6 Luna | 0.723 | 2 | 1.97 | 5 | Yes | Top 8 by both metrics |
| DeepL Next Gen | 0.721 | 3 | 1.94 | 2 | Yes | Top 8 by both metrics |
| Google Gemini 3.5 Flash | 0.720 | 4 | 1.96 | 3 | Yes | Top 8 by both metrics |
| Google Gemini 3.1 Pro Preview | 0.717 | 5 | 1.86 | 1 | Yes | Top 8 by both metrics |
| OpenAI GPT-5.5 | 0.717 | 6 | 2.04 | 9 | Yes | Top 8 by COMET-Kiwi |
| Google Translation LLM | 0.717 | 7 | 2.30 | 15 | Yes | Top 8 by COMET-Kiwi |
| z.ai GLM-5 | 0.714 | 8 | 2.05 | 10 | Yes | Top 8 by COMET-Kiwi |
| Anthropic Claude Fable 5 | 0.713 | 9 | 2.00 | 6 | Yes | Top 8 by MetricX |
| OpenAI GPT-4.1 | 0.711 | 10 | 2.02 | 7 | Yes | Top 8 by MetricX |
| Anthropic Claude Opus 4.8 | 0.710 | 11 | 2.02 | 8 | Yes | Top 8 by MetricX |
| Translated Lara 2 | 0.701 | 17 | 2.44 | 18 | Yes | Real-time place |
| Model | Solution complexity | COMET-Kiwi rank | COMET-Kiwi | MetricX rank | MetricX | Shortlist |
|---|---|---|---|---|---|---|
| xAI Grok 4.20 (reasoning) | Adjusted LLM | 1 | 0.724 | 4 | 1.96 | Top 8 by both metrics |
| OpenAI GPT-5.6 Luna | Adjusted LLM | 2 | 0.723 | 5 | 1.97 | Top 8 by both metrics |
| DeepL Next Gen | Off-the-shelf NMT | 3 | 0.721 | 2 | 1.94 | Top 8 by both metrics |
| Google Gemini 3.5 Flash | Adjusted LLM | 4 | 0.720 | 3 | 1.96 | Top 8 by both metrics |
| Google Gemini 3.1 Pro Preview | Adjusted LLM | 5 | 0.717 | 1 | 1.86 | Top 8 by both metrics |
| OpenAI GPT-5.5 | Adjusted LLM | 6 | 0.717 | 9 | 2.04 | Top 8 by COMET-Kiwi |
| Google Translation LLM | Adjusted NMT | 7 | 0.717 | 15 | 2.30 | Top 8 by COMET-Kiwi |
| z.ai GLM-5 | Adjusted LLM | 8 | 0.714 | 10 | 2.05 | Top 8 by COMET-Kiwi |
| Anthropic Claude Fable 5 | Adjusted LLM | 9 | 0.713 | 6 | 2.00 | Top 8 by MetricX |
| OpenAI GPT-4.1 | Adjusted LLM | 10 | 0.711 | 7 | 2.02 | Top 8 by MetricX |
| Anthropic Claude Opus 4.8 | Adjusted LLM | 11 | 0.710 | 8 | 2.02 | Top 8 by MetricX |
| DeepSeek V3.2 | Adjusted LLM | 12 | 0.709 | 13 | 2.17 | Not shortlisted |
| Anthropic Claude Sonnet 5 | Adjusted LLM | 13 | 0.709 | 12 | 2.16 | Not shortlisted |
| MoonshotAI Kimi K2.5 | Adjusted LLM | 14 | 0.706 | 19 | 2.48 | Not shortlisted |
| Mistral Large 3 | Adjusted LLM | 15 | 0.706 | 14 | 2.23 | Not shortlisted |
| Anthropic Claude Sonnet 4.6 | Adjusted LLM | 16 | 0.705 | 11 | 2.13 | Not shortlisted |
| Translated Lara 2 | Adjusted NMT | 17 | 0.701 | 18 | 2.44 | Real-time place |
| Tilde TildeOpen LLM engine | Adjusted NMT | 18 | 0.701 | 16 | 2.32 | Not shortlisted |
| Microsoft Translator | Off-the-shelf NMT | 19 | 0.698 | 17 | 2.34 | Not shortlisted |
| Google Cloud Advanced Translation API | Off-the-shelf NMT | 20 | 0.689 | 25 | 3.01 | Not shortlisted |
| OpenAI gpt-oss-120b | Adjusted LLM | 21 | 0.689 | 23 | 2.79 | Not shortlisted |
| Mistral Medium 3.1 | Adjusted LLM | 22 | 0.684 | 20 | 2.67 | Not shortlisted |
| ModernMT Enterprise Edition | Off-the-shelf NMT | 23 | 0.681 | 24 | 2.97 | Not shortlisted |
| Yandex Cloud Translate v2 | Off-the-shelf NMT | 24 | 0.681 | 21 | 2.77 | Not shortlisted |
| Anthropic Claude Haiku 4.5 | Adjusted LLM | 25 | 0.679 | 22 | 2.78 | Not shortlisted |
| Google Gemma 4 | Off-the-shelf LLM | 26 | 0.670 | 30 | 4.49 | Not shortlisted |
| Swiss AI Apertus-8B-2509 | Off-the-shelf LLM | 27 | 0.658 | 26 | 3.09 | Not shortlisted |
| MiniMax M2.5 | Adjusted LLM | 28 | 0.649 | 28 | 3.60 | Not shortlisted |
| Amazon Translate | Off-the-shelf NMT | 29 | 0.644 | 29 | 3.93 | Not shortlisted |
| Cohere Tiny Aya | Adjusted LLM | 30 | 0.640 | 27 | 3.57 | Not shortlisted |
| Alibaba Qwen3-235B-A22B | Off-the-shelf LLM | 31 | 0.581 | 31 | 5.36 | Not shortlisted |
| Cohere Command A Translate | Adjusted LLM | 32 | 0.520 | 32 | 8.01 | Not shortlisted |
| AI21 Jamba Large 1.7 | Adjusted LLM | 33 | 0.474 | 33 | 9.59 | Not shortlisted |
| OpenAI GPT-5.6 Sol | Adjusted LLM | — | 0.713 | — | 1.96 | Same model family as OpenAI GPT-5.6 Luna |
| OpenAI GPT-5.6 Terra | Adjusted LLM | — | 0.713 | — | 2.13 | Same model family as OpenAI GPT-5.6 Luna |
| xAI Grok 4.20 (non-reasoning) | Adjusted LLM | — | 0.709 | — | 2.01 | Same model family as xAI Grok 4.20 (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 33 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.
| Quality level | Cheapest solution that meets every requirement of the level | Cost per million characters |
|---|---|---|
| Basic | OpenAI GPT-5.6 Luna, three-step | $548 |
| Standard | OpenAI GPT-5.6 Luna, three-step | $548 |
| Advanced | Google Gemini 3.1 Pro Preview, three-step with the gender-neutrality agent | $951 |
| Model | US dollars per million characters | Class |
|---|---|---|
| OpenAI GPT-5.6 Luna | $0.49 | LLMs |
| xAI Grok 4.20 (reasoning) | $1.34 | LLMs |
| z.ai GLM-5 | $2.24 | LLMs |
| OpenAI GPT-4.1 | $3.34 | LLMs |
| Google Gemini 3.5 Flash | $4.12 | LLMs |
| Google Gemini 3.1 Pro Preview | $5.5 | LLMs |
| OpenAI GPT-5.5 | $12.19 | LLMs |
| Anthropic Claude Opus 4.8 | $16.69 | LLMs |
| Google Translation LLM | $21.94 | LLM-based MT engines |
| Translated Lara 2 | $24.99 | LLM-based MT engines |
| DeepL Next Gen | $27.5 | LLM-based MT engines |
| Anthropic Claude Fable 5 | $33.43 | LLMs |
| Model | Tokens per character | Class |
|---|---|---|
| Anthropic Claude Fable 5 | 0.523 | LLMs |
| Anthropic Claude Opus 4.8 | 0.523 | LLMs |
| Anthropic Claude Sonnet 5 | 0.522 | LLMs |
| Cohere Command A Translate | 0.452 | LLMs |
| Mistral Medium 3.1 | 0.449 | LLMs |
| Mistral Large 3 | 0.444 | LLMs |
| Anthropic Claude Haiku 4.5 | 0.441 | LLMs |
| AI21 Jamba Large 1.7 | 0.44 | LLMs |
| Anthropic Claude Sonnet 4.6 | 0.44 | LLMs |
| Alibaba Qwen3-235B-A22B | 0.402 | Open-weight LLMs |
| DeepSeek V3.2 | 0.399 | LLMs |
| MoonshotAI Kimi K2.5 | 0.391 | LLMs |
| z.ai GLM-5 | 0.39 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 0.37 | LLMs |
| xAI Grok 4.20 (reasoning) | 0.37 | LLMs |
| Google Gemini 3.1 Pro Preview | 0.364 | LLMs |
| Google Gemini 3.5 Flash | 0.364 | LLMs |
| Google Gemma 4 | 0.359 | Open-weight LLMs |
| MiniMax M2.5 | 0.353 | LLMs |
| Cohere Tiny Aya | 0.341 | LLMs |
| OpenAI GPT-4.1 | 0.324 | LLMs |
| OpenAI GPT-5.6 Luna | 0.323 | LLMs |
| OpenAI GPT-5.5 | 0.322 | LLMs |
| OpenAI GPT-5.6 Sol | 0.322 | LLMs |
| OpenAI GPT-5.6 Terra | 0.322 | LLMs |
| Quality level | Fastest solution that meets every requirement of the level | Seconds per 1,000 words |
|---|---|---|
| Basic | OpenAI GPT-5.6 Luna, three-step | 139 |
| Standard | OpenAI GPT-5.6 Luna, three-step | 139 |
| Advanced | OpenAI GPT-5.6 Luna, three-step with the gender-neutrality agent | 141 |
| Model | Seconds per 1,000 words | Class |
|---|---|---|
| Google Cloud Advanced Translation API | 0.1 | MT engines |
| Microsoft Translator | 0.2 | MT engines |
| ModernMT Enterprise Edition | 0.4 | MT engines |
| Google Translation LLM | 0.7 | LLM-based MT engines |
| DeepL Next Gen | 0.9 | LLM-based MT engines |
| Amazon Translate | 1.2 | MT engines |
| Translated Lara 2 | 1.6 | LLM-based MT engines |
| OpenAI gpt-oss-120b | 2.5 | Open-weight LLMs |
| Yandex Cloud Translate v2 | 3.2 | MT engines |
| Tilde TildeOpen LLM engine | 9.8 | LLM-based MT engines |
| OpenAI GPT-5.6 Luna | 17 | LLMs |
| Cohere Tiny Aya | 23 | LLMs |
| OpenAI GPT-4.1 | 33 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 34 | LLMs |
| Anthropic Claude Haiku 4.5 | 51 | LLMs |
| Swiss AI Apertus-8B-2509 | 51 | Open-weight LLMs |
| Mistral Medium 3.1 | 57 | LLMs |
| OpenAI GPT-5.6 Terra | 57 | LLMs |
| MiniMax M2.5 | 63 | LLMs |
| AI21 Jamba Large 1.7 | 65 | LLMs |
| DeepSeek V3.2 | 66 | LLMs |
| Cohere Command A Translate | 66 | LLMs |
| xAI Grok 4.20 (reasoning) | 75 | LLMs |
| Mistral Large 3 | 93 | LLMs |
| Anthropic Claude Sonnet 5 | 99 | LLMs |
| Anthropic Claude Sonnet 4.6 | 100 | LLMs |
| Anthropic Claude Fable 5 | 105 | LLMs |
| Google Gemini 3.5 Flash | 106 | LLMs |
| OpenAI GPT-5.6 Sol | 112 | LLMs |
| Google Gemini 3.1 Pro Preview | 115 | LLMs |
| Anthropic Claude Opus 4.8 | 126 | LLMs |
| OpenAI GPT-5.5 | 168 | LLMs |
| Alibaba Qwen3-235B-A22B | 181 | Open-weight LLMs |
| MoonshotAI Kimi K2.5 | 255 | LLMs |
| z.ai GLM-5 | 279 | LLMs |
| Google Gemma 4 | 280 | Open-weight LLMs |
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 Irish.
The real-time models in Irish are four MT engines, three LLM-based MT engines and the open-weight OpenAI gpt-oss-120b. Translations from the three on the shortlist, DeepL Next Gen, Google Translation LLM and Translated Lara 2, meet tone of voice. None of the three meet general translation quality or tag handling. The three-step solutions (translation + LQA + repair) that meet every Standard requirement take 139 to 415 seconds per 1,000 words, and the fastest that meets every Advanced requirement 141.
| The two best solutions | Solution | Sentences about people that are gender-neutral | Meets every Advanced requirement | Cost per million characters | Seconds per 1,000 words |
|---|---|---|---|---|---|
| Google Gemini 3.1 Pro Preview | three-step, with the gender-neutrality agent | 100% | ✓ | $951 | 320 |
| OpenAI GPT-5.6 Luna | three-step, with the gender-neutrality agent | 100% | ✓ | $963 | 141 |
For Irish we check the formal register, and gender-neutral wording through neutral nouns and rephrasing, with no default gender. The other options in Irish are a masculine default and paired forms. With the gender-neutrality agent, 100% of the sentences about people are gender-neutral in translations from the three-step solutions (translation + LQA + repair) on Google Gemini 3.1 Pro Preview and OpenAI GPT-5.6 Luna. Their weighted errors rise from 6.9 to 8.6 and from 13.8 to 17.5 per 1,000 words, still under the bar.
For the 12 shortlisted solutions in Irish, the median share of sentences on which the three LQA judges agree is 92.4%. With the LQA judges at that level of agreement, the Intento LQA error counts on this page carry weight. With both pairs of automatic metrics in agreement, so does the shortlist.
Leave feedback