The State of Translation Automation 2026 · Languages
What it takes to automate translation from English into Danish 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 Danish. |
| 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 Danish, 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.
Two one-step solutions meet every Standard requirement in Danish, and neither leads on errors, cost and time at once. Translations from OpenAI GPT-5.5 have 12.2 total errors per 1,000 words and no major error, against 13.8 and one major error from OpenAI GPT-5.6 Terra. The solution on OpenAI GPT-5.6 Terra costs $4.09 per million characters against $10.35, and takes 25 seconds per 1,000 words against 54. So Optimize for selects OpenAI GPT-5.5 for quality and OpenAI GPT-5.6 Terra for cost and latency.
Under the Standard requirements, a cost cap from $4.10 to $10.34 per million characters, or a time limit from 25 to under 54 seconds per 1,000 words, leaves only the solution on OpenAI GPT-5.6 Terra. A limit from 12.2 to under 13.8 total errors per 1,000 words leaves only the one on OpenAI GPT-5.5. None of the solutions under $4.09 per million characters or 24 seconds per 1,000 words meet even the Basic requirements. A limit of one step removes neither solution at any level.
The Advanced requirements do not change the cost, the time or the errors in Danish. Both solutions meet gender neutrality in one step, without the gender-neutrality agent. So for quality, cost and latency alike, Optimize for selects the same solution as under the Standard requirements.
OpenAI GPT-5.5
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.6 Terra
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.6 Terra
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.5
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.6 Terra
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.6 Terra
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.5
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.6 Terra
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.6 Terra
one-step LLM adjusted for Standard requirements
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 1: simple solution + backup | ★★ | one-step | OpenAI GPT-5.6 Terra | 13.8 | $4.09 | 25 |
| Standard | Tier 1: simple solution + backup | ★★ | one-step | OpenAI GPT-5.6 Terra | 13.8 | $4.09 | 25 |
| Advanced | Tier 1: simple solution + backup | ★★ | one-step | OpenAI GPT-5.6 Terra | 13.8 | $4.09 | 25 |
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 two one-step solutions meet every Basic, Standard and Advanced requirement in Danish, so a level up removes neither of them, and each level takes one step without the gender-neutrality agent. General translation quality, the one Basic requirement, is where the other nine one-step solutions with a conclusive Intento LQA result fall short. Under all three levels, Optimize for selects OpenAI GPT-5.5 for quality, with 12.2 total errors per 1,000 words, and OpenAI GPT-5.6 Terra for cost and latency.
| Model | US dollars per million characters | Weighted errors per 1,000 words | Group |
|---|---|---|---|
| Mistral Large 3 | 0.74 | 22.7 | Misses a Standard requirement |
| DeepSeek V3.2 | 0.9 | 32.6 | Misses a Standard requirement |
| MoonshotAI Kimi K2.5 | 1.23 | 33.9 | Misses a Standard requirement |
| OpenAI GPT-4.1 | 2.86 | 25 | Misses a Standard requirement |
| OpenAI GPT-5.6 Terra | 4.09 | 13.8 | Meets every Standard requirement |
| Google Gemini 3.1 Pro Preview | 4.36 | 23.4 | Misses a Standard requirement |
| Microsoft Translator | 10 | 56.7 | Misses a Standard requirement |
| OpenAI GPT-5.5 | 10.35 | 12.2 | Meets every Standard requirement |
| Anthropic Claude Opus 4.8 | 13.76 | 23.4 | Misses a Standard requirement |
| Translated Lara 2 | 24.99 | 31.6 | Misses a Standard requirement |
| DeepL Next Gen | 27.5 | 22.7 | Misses a Standard requirement |
In Danish, the three one-step solutions that cost more than the one on OpenAI GPT-5.5, at $10.35 per million characters, all have more weighted errors per 1,000 words. Translations from DeepL Next Gen, the dearest at $27.50 per million characters, have 22.7 weighted errors per 1,000 words, as many as those from Mistral Large 3, the cheapest at $0.74. The two solutions that meet every Standard requirement are on LLMs, OpenAI GPT-5.6 Terra and OpenAI GPT-5.5, with 13.8 and 12.2 weighted errors per 1,000 words.
| Requirement | Median errors per 1,000 words |
|---|---|
| General translation quality | 23.4 |
| Tone of voice | 0 |
| Tag handling | 0 |
| Gender neutrality | 0 |
The requirements with countable errors. Every solution with its result on each requirement: the explorer.
General translation quality is the Standard requirement the most one-step solutions in Danish miss, and five of the nine that miss it meet tone of voice, tag handling and formatting. Translations from Mistral Large 3, Google Gemini 3.1 Pro Preview, Anthropic Claude Opus 4.8 and OpenAI GPT-4.1 have 22.7 to 25.0 weighted errors per 1,000 words, over the bar. Those from DeepSeek V3.2 have a critical error. With two-step solutions (translation + post-edit) or three-step solutions (translation + LQA + repair), Danish stays in tier 1.
| Model | Weighted errors per 1,000 words of the shortlisted one-step solutions in Danish | Class |
|---|---|---|
| Microsoft Translator | 56.7 | MT engines |
| MoonshotAI Kimi K2.5 | 33.9 | LLMs |
| DeepSeek V3.2 | 32.6 | LLMs |
| Translated Lara 2 | 31.6 | LLM-based MT engines |
| OpenAI GPT-4.1 | 25 | LLMs |
| Anthropic Claude Opus 4.8 | 23.4 | LLMs |
| Google Gemini 3.1 Pro Preview | 23.4 | LLMs |
| Mistral Large 3 | 22.7 | LLMs |
| DeepL Next Gen | 22.7 | LLM-based MT engines |
| OpenAI GPT-5.6 Terra | 13.8 | LLMs |
| OpenAI GPT-5.5 | 12.2 | LLMs |
No translation into Danish: SpeakLeash Bielik-11B-v3.0-Instruct (wrong language), Widn AI Tower Anthill (cut off part-way, wrong language), Widn AI Tower Vesuvius (cut off part-way).
Translations from the eight shortlisted LLMs with a conclusive Intento LQA result have 12.2 to 33.9 weighted errors per 1,000 words in Danish, and all eight support the language by our measured criteria. Those from the LLM-based MT engines DeepL Next Gen and Translated Lara 2 have 22.7 and 31.6 weighted errors per 1,000 words, and those from the MT engine Microsoft Translator 56.7. Translations from DeepL Next Gen have 1.86 times as many weighted errors as those from OpenAI GPT-5.5.
| Model | COMET-Kiwi | Rank on COMET-Kiwi | MetricX | Rank on MetricX | Shortlisted | Shortlist place |
|---|---|---|---|---|---|---|
| z.ai GLM-5 | 0.790 | 9 | 1.13 | 9 | No | Not shortlisted |
| Anthropic Claude Fable 5 | 0.786 | 11 | 1.30 | 20 | No | Not shortlisted |
| Widn AI Tower Sugarloaf | 0.785 | 14 | 1.30 | 18 | No | Not shortlisted |
| Google Gemma 4 | 0.785 | 15 | 1.28 | 16 | No | Not shortlisted |
| Swiss AI Apertus-8B-2509 | 0.785 | 16 | 1.17 | 10 | No | Not shortlisted |
| Cohere Command A Translate | 0.784 | 17 | 1.21 | 12 | No | Not shortlisted |
| xAI Grok 4.20 (reasoning) | 0.784 | 18 | 1.32 | 21 | No | Not shortlisted |
| Google Gemini 3.5 Flash | 0.784 | 19 | 1.36 | 24 | No | Not shortlisted |
| Tilde TildeOpen LLM engine | 0.783 | 20 | 1.37 | 25 | No | Not shortlisted |
| Anthropic Claude Haiku 4.5 | 0.783 | 22 | 1.21 | 13 | No | Not shortlisted |
| Anthropic Claude Sonnet 4.6 | 0.783 | 23 | 1.21 | 14 | No | Not shortlisted |
| Anthropic Claude Sonnet 5 | 0.783 | 24 | 1.29 | 17 | No | Not shortlisted |
| Google Translation LLM | 0.783 | 25 | 1.45 | 28 | No | Not shortlisted |
| Alibaba Qwen3-235B-A22B | 0.780 | 26 | 1.36 | 23 | No | Not shortlisted |
| ModernMT Enterprise Edition | 0.778 | 27 | 1.30 | 19 | No | Not shortlisted |
| SYSTRAN PNMT API | 0.775 | 28 | 1.34 | 22 | No | Not shortlisted |
| Amazon Translate | 0.770 | 29 | 1.47 | 29 | No | Not shortlisted |
| Supertext LLM | 0.770 | 30 | 1.51 | 30 | No | Not shortlisted |
| MiniMax M2.5 | 0.767 | 31 | 1.61 | 32 | No | Not shortlisted |
| AI21 Jamba Large 1.7 | 0.767 | 32 | 1.56 | 31 | No | Not shortlisted |
| Oracle Cloud Language Service | 0.756 | 33 | 1.72 | 33 | No | Not shortlisted |
| Google Cloud Advanced Translation API | 0.756 | 34 | 2.14 | 36 | No | Not shortlisted |
| Cohere Tiny Aya | 0.753 | 35 | 1.81 | 34 | No | Not shortlisted |
| Yandex Cloud Translate v2 | 0.743 | 36 | 1.97 | 35 | No | Not shortlisted |
| OpenAI gpt-oss-120b | 0.722 | 37 | 2.80 | 39 | No | Not shortlisted |
| AppTek NMT | 0.714 | 38 | 2.66 | 38 | No | Not shortlisted |
| Alibaba Machine Translation API | 0.696 | 39 | 2.66 | 37 | No | Not shortlisted |
| DeepL Next Gen | 0.800 | 1 | 0.99 | 1 | Yes | Top 8 by both metrics |
| OpenAI GPT-4.1 | 0.797 | 2 | 1.12 | 6 | Yes | Top 8 by both metrics |
| OpenAI GPT-5.6 Terra | 0.797 | 3 | 1.07 | 3 | Yes | Top 8 by both metrics |
| Anthropic Claude Opus 4.8 | 0.796 | 4 | 1.20 | 11 | Yes | Top 8 by COMET-Kiwi |
| Google Gemini 3.1 Pro Preview | 0.795 | 5 | 1.08 | 5 | Yes | Top 8 by both metrics |
| OpenAI GPT-5.5 | 0.795 | 6 | 1.07 | 2 | Yes | Top 8 by both metrics |
| Mistral Large 3 | 0.791 | 7 | 1.12 | 7 | Yes | Top 8 by both metrics |
| DeepSeek V3.2 | 0.791 | 8 | 1.25 | 15 | Yes | Top 8 by COMET-Kiwi |
| MoonshotAI Kimi K2.5 | 0.788 | 10 | 1.13 | 8 | Yes | Top 8 by MetricX |
| Mistral Medium 3.1 | 0.786 | 12 | 1.08 | 4 | Yes | Top 8 by MetricX |
| Translated Lara 2 | 0.786 | 13 | 1.38 | 26 | Yes | Real-time place |
| Microsoft Translator | 0.783 | 21 | 1.39 | 27 | Yes | Real-time place |
| Model | Solution complexity | COMET-Kiwi rank | COMET-Kiwi | MetricX rank | MetricX | Shortlist |
|---|---|---|---|---|---|---|
| DeepL Next Gen | Off-the-shelf NMT | 1 | 0.800 | 1 | 0.99 | Top 8 by both metrics |
| OpenAI GPT-4.1 | Adjusted LLM | 2 | 0.797 | 6 | 1.12 | Top 8 by both metrics |
| OpenAI GPT-5.6 Terra | Adjusted LLM | 3 | 0.797 | 3 | 1.07 | Top 8 by both metrics |
| Anthropic Claude Opus 4.8 | Adjusted LLM | 4 | 0.796 | 11 | 1.20 | Top 8 by COMET-Kiwi |
| Google Gemini 3.1 Pro Preview | Adjusted LLM | 5 | 0.795 | 5 | 1.08 | Top 8 by both metrics |
| OpenAI GPT-5.5 | Adjusted LLM | 6 | 0.795 | 2 | 1.07 | Top 8 by both metrics |
| Mistral Large 3 | Adjusted LLM | 7 | 0.791 | 7 | 1.12 | Top 8 by both metrics |
| DeepSeek V3.2 | Adjusted LLM | 8 | 0.791 | 15 | 1.25 | Top 8 by COMET-Kiwi |
| z.ai GLM-5 | Adjusted LLM | 9 | 0.790 | 9 | 1.13 | Not shortlisted |
| MoonshotAI Kimi K2.5 | Adjusted LLM | 10 | 0.788 | 8 | 1.13 | Top 8 by MetricX |
| Anthropic Claude Fable 5 | Adjusted LLM | 11 | 0.786 | 20 | 1.30 | Not shortlisted |
| Mistral Medium 3.1 | Adjusted LLM | 12 | 0.786 | 4 | 1.08 | Top 8 by MetricX |
| Translated Lara 2 | Adjusted NMT | 13 | 0.786 | 26 | 1.38 | Real-time place |
| Widn AI Tower Sugarloaf | Adjusted NMT | 14 | 0.785 | 18 | 1.30 | Not shortlisted |
| Google Gemma 4 | Adjusted LLM | 15 | 0.785 | 16 | 1.28 | Not shortlisted |
| Swiss AI Apertus-8B-2509 | Adjusted LLM | 16 | 0.785 | 10 | 1.17 | Not shortlisted |
| Cohere Command A Translate | Adjusted LLM | 17 | 0.784 | 12 | 1.21 | Not shortlisted |
| xAI Grok 4.20 (reasoning) | Adjusted LLM | 18 | 0.784 | 21 | 1.32 | Not shortlisted |
| Google Gemini 3.5 Flash | Adjusted LLM | 19 | 0.784 | 24 | 1.36 | Not shortlisted |
| Tilde TildeOpen LLM engine | Adjusted NMT | 20 | 0.783 | 25 | 1.37 | Not shortlisted |
| Microsoft Translator | Off-the-shelf NMT | 21 | 0.783 | 27 | 1.39 | Real-time place |
| Anthropic Claude Haiku 4.5 | Adjusted LLM | 22 | 0.783 | 13 | 1.21 | Not shortlisted |
| Anthropic Claude Sonnet 4.6 | Adjusted LLM | 23 | 0.783 | 14 | 1.21 | Not shortlisted |
| Anthropic Claude Sonnet 5 | Adjusted LLM | 24 | 0.783 | 17 | 1.29 | Not shortlisted |
| Google Translation LLM | Adjusted NMT | 25 | 0.783 | 28 | 1.45 | Not shortlisted |
| Alibaba Qwen3-235B-A22B | Adjusted LLM | 26 | 0.780 | 23 | 1.36 | Not shortlisted |
| ModernMT Enterprise Edition | Off-the-shelf NMT | 27 | 0.778 | 19 | 1.30 | Not shortlisted |
| SYSTRAN PNMT API | Off-the-shelf NMT | 28 | 0.775 | 22 | 1.34 | Not shortlisted |
| Amazon Translate | Off-the-shelf NMT | 29 | 0.770 | 29 | 1.47 | Not shortlisted |
| Supertext LLM | Off-the-shelf NMT | 30 | 0.770 | 30 | 1.51 | Not shortlisted |
| MiniMax M2.5 | Adjusted LLM | 31 | 0.767 | 32 | 1.61 | Not shortlisted |
| AI21 Jamba Large 1.7 | Adjusted LLM | 32 | 0.767 | 31 | 1.56 | Not shortlisted |
| Oracle Cloud Language Service | Off-the-shelf NMT | 33 | 0.756 | 33 | 1.72 | Not shortlisted |
| Google Cloud Advanced Translation API | Off-the-shelf NMT | 34 | 0.756 | 36 | 2.14 | Not shortlisted |
| Cohere Tiny Aya | Off-the-shelf LLM | 35 | 0.753 | 34 | 1.81 | Not shortlisted |
| Yandex Cloud Translate v2 | Off-the-shelf NMT | 36 | 0.743 | 35 | 1.97 | Not shortlisted |
| OpenAI gpt-oss-120b | Adjusted LLM | 37 | 0.722 | 39 | 2.80 | Not shortlisted |
| AppTek NMT | Off-the-shelf NMT | 38 | 0.714 | 38 | 2.66 | Not shortlisted |
| Alibaba Machine Translation API | Off-the-shelf NMT | 39 | 0.696 | 37 | 2.66 | Not shortlisted |
| OpenAI GPT-5.6 Luna | Adjusted LLM | — | 0.794 | — | 1.13 | Same model family as OpenAI GPT-5.6 Terra |
| OpenAI GPT-5.6 Sol | Adjusted LLM | — | 0.795 | — | 1.05 | Same model family as OpenAI GPT-5.6 Terra |
| xAI Grok 4.20 (non-reasoning) | Adjusted LLM | — | 0.780 | — | 1.38 | 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 39 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.
Four solutions came onto the shortlist through one metric alone. Anthropic Claude Opus 4.8 and DeepSeek V3.2 are in the top 8 of COMET-Kiwi but 11th and 15th on MetricX, and MoonshotAI Kimi K2.5 and Mistral Medium 3.1 in the top 8 of MetricX only. By Intento LQA, translations from the first three have 23.4 to 33.9 weighted errors per 1,000 words, over the bar, and the solution on Mistral Medium 3.1 has no conclusive result. Nearest the cut in the tail is z.ai GLM-5, 9th on both metrics.
| Quality level | Cheapest solution that meets every requirement of the level | Cost per million characters |
|---|---|---|
| Basic | OpenAI GPT-5.6 Terra, one-step LLM adjusted for Standard requirements | $4.09 |
| Standard | OpenAI GPT-5.6 Terra, one-step LLM adjusted for Standard requirements | $4.09 |
| Advanced | OpenAI GPT-5.6 Terra, one-step LLM adjusted for Standard requirements | $4.09 |
| Model | US dollars per million characters | Class |
|---|---|---|
| Mistral Large 3 | $0.74 | LLMs |
| DeepSeek V3.2 | $0.9 | LLMs |
| Mistral Medium 3.1 | $0.92 | LLMs |
| MoonshotAI Kimi K2.5 | $1.23 | LLMs |
| OpenAI GPT-4.1 | $2.87 | LLMs |
| OpenAI GPT-5.6 Terra | $4.09 | LLMs |
| Google Gemini 3.1 Pro Preview | $4.36 | LLMs |
| Microsoft Translator | $10 | MT engines |
| OpenAI GPT-5.5 | $10.35 | LLMs |
| Anthropic Claude Opus 4.8 | $13.76 | LLMs |
| Translated Lara 2 | $24.99 | LLM-based MT engines |
| DeepL Next Gen | $27.5 | LLM-based MT engines |
| Model | Tokens per character | Class |
|---|---|---|
| Anthropic Claude Opus 4.8 | 0.451 | LLMs |
| Anthropic Claude Fable 5 | 0.448 | LLMs |
| Anthropic Claude Sonnet 5 | 0.447 | LLMs |
| Mistral Medium 3.1 | 0.388 | LLMs |
| Mistral Large 3 | 0.381 | LLMs |
| AI21 Jamba Large 1.7 | 0.369 | LLMs |
| Cohere Command A Translate | 0.342 | LLMs |
| MoonshotAI Kimi K2.5 | 0.339 | LLMs |
| Anthropic Claude Haiku 4.5 | 0.336 | LLMs |
| Anthropic Claude Sonnet 4.6 | 0.336 | LLMs |
| Widn AI Tower Sugarloaf | 0.334 | LLM-based MT engines |
| Alibaba Qwen3-235B-A22B | 0.333 | Open-weight LLMs |
| DeepSeek V3.2 | 0.331 | LLMs |
| z.ai GLM-5 | 0.317 | LLMs |
| Google Gemini 3.5 Flash | 0.302 | LLMs |
| xAI Grok 4.20 (reasoning) | 0.302 | LLMs |
| Google Gemini 3.1 Pro Preview | 0.301 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 0.301 | LLMs |
| Google Gemma 4 | 0.3 | Open-weight LLMs |
| OpenAI GPT-4.1 | 0.286 | LLMs |
| OpenAI GPT-5.6 Terra | 0.286 | LLMs |
| OpenAI GPT-5.5 | 0.285 | LLMs |
| OpenAI GPT-5.6 Luna | 0.285 | LLMs |
| OpenAI GPT-5.6 Sol | 0.285 | LLMs |
| MiniMax M2.5 | 0.282 | LLMs |
| Cohere Tiny Aya | 0.263 | LLMs |
The tokenizers produce a median 0.32 tokens per character of the Danish translation, 1.3 times their count for English. Those of OpenAI GPT-5.5 and OpenAI GPT-5.6 Terra produce 0.285 and 0.286 tokens per character, almost the same, while the two solutions cost $10.35 and $4.09 per million characters. Under all three levels, the cheapest solution that meets every requirement is the same, the one-step solution on OpenAI GPT-5.6 Terra at $4.09 per million characters, without the gender-neutrality agent.
| Quality level | Fastest solution that meets every requirement of the level | Seconds per 1,000 words |
|---|---|---|
| Basic | OpenAI GPT-5.6 Terra, one-step LLM adjusted for Standard requirements | 25 |
| Standard | OpenAI GPT-5.6 Terra, one-step LLM adjusted for Standard requirements | 25 |
| Advanced | OpenAI GPT-5.6 Terra, one-step LLM adjusted for Standard requirements | 25 |
| Model | Seconds per 1,000 words | Class |
|---|---|---|
| Google Cloud Advanced Translation API | 0.1 | MT engines |
| Microsoft Translator | 0.2 | MT engines |
| Google Translation LLM | 0.4 | LLM-based MT engines |
| ModernMT Enterprise Edition | 0.4 | MT engines |
| DeepL Next Gen | 0.6 | LLM-based MT engines |
| Translated Lara 2 | 1.1 | LLM-based MT engines |
| Amazon Translate | 1.3 | MT engines |
| SYSTRAN PNMT API | 1.7 | MT engines |
| OpenAI gpt-oss-120b | 1.9 | Open-weight LLMs |
| Yandex Cloud Translate v2 | 2.6 | MT engines |
| Oracle Cloud Language Service | 4.4 | MT engines |
| Tilde TildeOpen LLM engine | 7.4 | LLM-based MT engines |
| Google Gemini 3.5 Flash | 14 | LLMs |
| Supertext LLM | 16 | LLM-based MT engines |
| Cohere Tiny Aya | 17 | LLMs |
| OpenAI GPT-5.6 Luna | 17 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 17 | LLMs |
| OpenAI GPT-4.1 | 22 | LLMs |
| Alibaba Machine Translation API | 23 | MT engines |
| OpenAI GPT-5.6 Terra | 25 | LLMs |
| Anthropic Claude Haiku 4.5 | 28 | LLMs |
| Mistral Medium 3.1 | 36 | LLMs |
| MiniMax M2.5 | 40 | LLMs |
| DeepSeek V3.2 | 43 | LLMs |
| Cohere Command A Translate | 47 | LLMs |
| Anthropic Claude Sonnet 5 | 47 | LLMs |
| Swiss AI Apertus-8B-2509 | 48 | Open-weight LLMs |
| AI21 Jamba Large 1.7 | 49 | LLMs |
| OpenAI GPT-5.6 Sol | 49 | LLMs |
| Widn AI Tower Sugarloaf | 50 | LLM-based MT engines |
| Mistral Large 3 | 51 | LLMs |
| Anthropic Claude Fable 5 | 53 | LLMs |
| OpenAI GPT-5.5 | 54 | LLMs |
| Anthropic Claude Sonnet 4.6 | 55 | LLMs |
| xAI Grok 4.20 (reasoning) | 57 | LLMs |
| Anthropic Claude Opus 4.8 | 61 | LLMs |
| Google Gemini 3.1 Pro Preview | 87 | LLMs |
| AppTek NMT | 88 | MT engines |
| Alibaba Qwen3-235B-A22B | 112 | Open-weight LLMs |
| Google Gemma 4 | 135 | Open-weight LLMs |
| z.ai GLM-5 | 195 | LLMs |
| MoonshotAI Kimi K2.5 | 238 | 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 Danish.
Five of the nine real-time models in Danish are MT engines, three are LLM-based MT engines and one is the open-weight OpenAI gpt-oss-120b. The three on the shortlist, on Microsoft Translator, DeepL Next Gen and Translated Lara 2, miss formatting and have 22.7 to 56.7 weighted errors per 1,000 words, over the bar for general translation quality.
| 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 |
|---|---|---|---|---|---|
| OpenAI GPT-5.5 | one-step | 100% | ✓ | $10.35 | 54 |
| OpenAI GPT-5.6 Terra | one-step | 100% | ✓ | $4.09 | 25 |
For Danish we check the formal register, and gender-neutral wording through rephrasing, plurals and neutral nouns, with no default gender. The other options in the language are the han/hun split, generic han and rephrasing with no standard neutral pronoun. Translations from all 11 one-step solutions with a conclusive Intento LQA result have 100% of the sentences in the formal register. In the translations from both of the two best solutions, 100% of the sentences about people are gender-neutral, without the gender-neutrality agent.
The three LQA judges agree on a median of 96.3% of the sentences of the 12 shortlisted solutions in Danish. Both pairs of automatic metrics agree above the threshold, and neither result is borderline. So the Intento LQA error counts and the shortlist ranks on this page rest on measurement we can rely on.
Leave feedback