The State of Translation Automation 2026 · Languages
What it takes to automate translation from English into Slovak 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 Slovak. |
| Tone of voice | The formal register (vy (vykanie)) 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 Slovak, 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 11 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. 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.
Neither of the two solutions that meet every Standard requirement is ahead on errors, time and cost at once. The solution on Anthropic Claude Fable 5 has 16.0 total errors per 1,000 words and takes 66 seconds per 1,000 words, at $29.37 per million characters. The solution on Google Gemini 3.1 Pro Preview has 18.3 total errors per 1,000 words and takes 87 seconds per 1,000 words, at $4.66 per million characters. Optimize for selects Anthropic Claude Fable 5 for quality and latency, and Google Gemini 3.1 Pro Preview for cost.
Under the Standard requirements, a cost cap below $29.37 per million characters leaves the solution on Google Gemini 3.1 Pro Preview alone, and a cap below $4.66 leaves none. A time limit below 87 seconds per 1,000 words leaves the solution on Anthropic Claude Fable 5 alone, and a limit of 65 seconds or less leaves none. None of the three real-time solutions (<2 sec) on the shortlist, on Google Translation LLM, DeepL Next Gen and Translated Lara 2, meet general translation quality.
Under the Advanced requirements, the three-step solution (translation + LQA + repair) on Anthropic Claude Fable 5 with the gender-neutrality agent has 5.3 total errors per 1,000 words, against 16.0 in one step. It takes 220 seconds per 1,000 words at $1,027 per million characters. The two-step solution (translation + post-edit) on Google Gemini 3.1 Pro Preview with the agent has 17.8 total errors per 1,000 words, against 18.3 in one step. It takes 108 seconds per 1,000 words at $4.90 per million characters.
Anthropic Claude Fable 5
one-step LLM adjusted for Standard requirements
Google Gemini 3.5 Flash
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.6 Terra
one-step LLM adjusted for Standard requirements
Anthropic Claude Fable 5
one-step LLM adjusted for Standard requirements
Google Gemini 3.1 Pro Preview
one-step LLM adjusted for Standard requirements
Anthropic Claude Fable 5
one-step LLM adjusted for Standard requirements
Anthropic Claude Fable 5
three-step with the gender-neutrality agent
Google Gemini 3.1 Pro Preview
two-step with the gender-neutrality agent
Google Gemini 3.1 Pro Preview
two-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 1: simple solution + backup | ★×4 | one-step | Google Gemini 3.5 Flash | 19.3 | $3.46 | 58 |
| Standard | Tier 1: simple solution + backup | ★★ | one-step | Google Gemini 3.1 Pro Preview | 18.3 | $4.66 | 87 |
| Advanced | Tier 2: two or more solutions, none simple | ☆☆ | with the gender-neutrality agent | Google Gemini 3.1 Pro Preview | 17.8 | $4.90 | 108 |
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.
Of the four solutions that meet the Basic requirements, the Standard ones remove those on OpenAI GPT-5.6 Terra for tag handling and Google Gemini 3.5 Flash for formatting. The Advanced requirements take a three-step solution (translation + LQA + repair) on Anthropic Claude Fable 5 or a two-step solution (translation + post-edit) on Google Gemini 3.1 Pro Preview, both with the gender-neutrality agent. Under the Basic requirements, Optimize for selects Google Gemini 3.5 Flash for cost and OpenAI GPT-5.6 Terra for latency.
| Model | US dollars per million characters | Weighted errors per 1,000 words | Group |
|---|---|---|---|
| MoonshotAI Kimi K2.5 | 1.37 | 29.7 | Misses a Standard requirement |
| z.ai GLM-5 | 2.08 | 39.6 | Misses a Standard requirement |
| OpenAI GPT-4.1 | 3.18 | 28.3 | Misses a Standard requirement |
| Google Gemini 3.5 Flash | 3.46 | 16.5 | Misses a Standard requirement |
| OpenAI GPT-5.6 Terra | 4.59 | 11.9 | Misses a Standard requirement |
| Google Gemini 3.1 Pro Preview | 4.66 | 15.5 | Meets every Standard requirement |
| Anthropic Claude Sonnet 4.6 | 6.98 | 33.3 | Misses a Standard requirement |
| Google Translation LLM | 20.62 | 57.4 | Misses a Standard requirement |
| Translated Lara 2 | 24.99 | 28.3 | Misses a Standard requirement |
| DeepL Next Gen | 27.5 | 29 | Misses a Standard requirement |
| Anthropic Claude Fable 5 | 29.37 | 12.9 | Meets every Standard requirement |
Among the shortlisted one-step solutions, a higher price does not always come with fewer errors. None of the three LLM-based MT engines, at $20.62 to $27.50 per million characters, meet general translation quality. The solution on OpenAI GPT-5.6 Terra costs $4.59 per million characters and has 11.9 weighted errors per 1,000 words, against 12.9 for the solution on Anthropic Claude Fable 5 at $29.37. The two cheapest, on MoonshotAI Kimi K2.5 and z.ai GLM-5, have 29.7 and 39.6 weighted errors per 1,000 words.
| Requirement | Median errors per 1,000 words |
|---|---|
| General translation quality | 28.3 |
| Gender neutrality | 3.2 |
| Tone of voice | 0 |
| Tag handling | 0 |
The requirements with countable errors. Every solution with its result on each requirement: the explorer.
Translations from three LLMs come close on tag handling. Those from MoonshotAI Kimi K2.5 have 99.6% of the tags right, and those from OpenAI GPT-5.6 Terra and Anthropic Claude Sonnet 4.6 have 99.2%. Of the LLM-based MT engines, translations from Translated Lara 2 have every tag right, against 95.1% for Google Translation LLM and 93.3% for DeepL Next Gen. One step is enough here, so two-step solutions (translation + post-edit) and three-step solutions (translation + LQA + repair) leave Slovak in tier 1 for the Standard requirements.
| Model | Weighted errors per 1,000 words of the shortlisted one-step solutions in Slovak | Class |
|---|---|---|
| Google Translation LLM | 57.4 | LLM-based MT engines |
| z.ai GLM-5 | 39.6 | LLMs |
| Anthropic Claude Sonnet 4.6 | 33.3 | LLMs |
| MoonshotAI Kimi K2.5 | 29.7 | LLMs |
| DeepL Next Gen | 29 | LLM-based MT engines |
| OpenAI GPT-4.1 | 28.3 | LLMs |
| Translated Lara 2 | 28.3 | LLM-based MT engines |
| Google Gemini 3.5 Flash | 16.5 | LLMs |
| Google Gemini 3.1 Pro Preview | 15.5 | LLMs |
| Anthropic Claude Fable 5 | 12.9 | LLMs |
| OpenAI GPT-5.6 Terra | 11.9 | LLMs |
No translation into Slovak: SpeakLeash Bielik-11B-v3.0-Instruct (wrong language), Widn AI Tower Anthill (far too short, wrong language), Widn AI Tower Vesuvius (far too short, wrong language).
The shortlist for Slovak holds eight LLMs and three LLM-based MT engines. Translations from the LLMs have 11.9 to 39.6 weighted errors per 1,000 words, and translations from the LLM-based MT engines 28.3 to 57.4. All eight shortlisted LLMs support Slovak by our measured criteria, and four of them meet general translation quality. None of the MT engines that are not LLM-based made the shortlist. None of the open-weight LLMs we got translations from made it either.
| Model | COMET-Kiwi | Rank on COMET-Kiwi | MetricX | Rank on MetricX | Shortlisted | Shortlist place |
|---|---|---|---|---|---|---|
| xAI Grok 4.20 (reasoning) | 0.813 | 9 | 1.79 | 13 | No | Not shortlisted |
| Anthropic Claude Sonnet 5 | 0.812 | 10 | 1.78 | 12 | No | Not shortlisted |
| Anthropic Claude Opus 4.8 | 0.810 | 12 | 1.74 | 9 | No | Not shortlisted |
| OpenAI GPT-5.5 | 0.810 | 13 | 1.87 | 15 | No | Not shortlisted |
| Cohere Command A Translate | 0.809 | 15 | 1.87 | 16 | No | Not shortlisted |
| DeepSeek V3.2 | 0.807 | 16 | 1.96 | 17 | No | Not shortlisted |
| Google Gemma 4 | 0.806 | 18 | 2.02 | 21 | No | Not shortlisted |
| Anthropic Claude Haiku 4.5 | 0.804 | 19 | 2.08 | 24 | No | Not shortlisted |
| Tilde TildeOpen LLM engine | 0.804 | 20 | 2.05 | 23 | No | Not shortlisted |
| Mistral Large 3 | 0.803 | 21 | 1.80 | 14 | No | Not shortlisted |
| Microsoft Translator | 0.797 | 22 | 2.02 | 22 | No | Not shortlisted |
| OpenAI gpt-oss-120b | 0.797 | 23 | 2.01 | 18 | No | Not shortlisted |
| Mistral Medium 3.1 | 0.795 | 24 | 2.01 | 20 | No | Not shortlisted |
| MiniMax M2.5 | 0.793 | 25 | 2.27 | 27 | No | Not shortlisted |
| ModernMT Enterprise Edition | 0.792 | 26 | 2.01 | 19 | No | Not shortlisted |
| Supertext LLM | 0.791 | 27 | 2.17 | 26 | No | Not shortlisted |
| Swiss AI Apertus-8B-2509 | 0.785 | 28 | 2.62 | 30 | No | Not shortlisted |
| Google Cloud Advanced Translation API | 0.784 | 29 | 2.68 | 31 | No | Not shortlisted |
| SYSTRAN PNMT API | 0.765 | 30 | 2.74 | 32 | No | Not shortlisted |
| AI21 Jamba Large 1.7 | 0.761 | 31 | 2.55 | 28 | No | Not shortlisted |
| Alibaba Qwen3-235B-A22B | 0.760 | 32 | 2.61 | 29 | No | Not shortlisted |
| Yandex Cloud Translate v2 | 0.756 | 33 | 2.96 | 34 | No | Not shortlisted |
| Cohere Tiny Aya | 0.756 | 34 | 2.84 | 33 | No | Not shortlisted |
| Amazon Translate | 0.752 | 35 | 3.08 | 35 | No | Not shortlisted |
| Alibaba Machine Translation API | 0.663 | 36 | 5.40 | 36 | No | Not shortlisted |
| AppTek NMT | 0.603 | 37 | 7.36 | 37 | No | Not shortlisted |
| DeepL Next Gen | 0.824 | 1 | 1.54 | 2 | Yes | Top 8 by both metrics |
| Google Gemini 3.1 Pro Preview | 0.819 | 2 | 1.70 | 6 | Yes | Top 8 by both metrics |
| OpenAI GPT-5.6 Terra | 0.819 | 3 | 1.51 | 1 | Yes | Top 8 by both metrics |
| Google Gemini 3.5 Flash | 0.818 | 4 | 1.71 | 7 | Yes | Top 8 by both metrics |
| Anthropic Claude Sonnet 4.6 | 0.817 | 5 | 1.65 | 3 | Yes | Top 8 by both metrics |
| z.ai GLM-5 | 0.815 | 6 | 1.69 | 5 | Yes | Top 8 by both metrics |
| MoonshotAI Kimi K2.5 | 0.814 | 7 | 1.76 | 10 | Yes | Top 8 by COMET-Kiwi |
| Anthropic Claude Fable 5 | 0.814 | 8 | 1.73 | 8 | Yes | Top 8 by both metrics |
| OpenAI GPT-4.1 | 0.810 | 11 | 1.67 | 4 | Yes | Top 8 by MetricX |
| Translated Lara 2 | 0.809 | 14 | 1.77 | 11 | Yes | Real-time place |
| Google Translation LLM | 0.806 | 17 | 2.09 | 25 | Yes | Real-time place |
| Model | Solution complexity | COMET-Kiwi rank | COMET-Kiwi | MetricX rank | MetricX | Shortlist |
|---|---|---|---|---|---|---|
| DeepL Next Gen | Off-the-shelf NMT | 1 | 0.824 | 2 | 1.54 | Top 8 by both metrics |
| Google Gemini 3.1 Pro Preview | Adjusted LLM | 2 | 0.819 | 6 | 1.70 | Top 8 by both metrics |
| OpenAI GPT-5.6 Terra | Adjusted LLM | 3 | 0.819 | 1 | 1.51 | Top 8 by both metrics |
| Google Gemini 3.5 Flash | Adjusted LLM | 4 | 0.818 | 7 | 1.71 | Top 8 by both metrics |
| Anthropic Claude Sonnet 4.6 | Adjusted LLM | 5 | 0.817 | 3 | 1.65 | Top 8 by both metrics |
| z.ai GLM-5 | Adjusted LLM | 6 | 0.815 | 5 | 1.69 | Top 8 by both metrics |
| MoonshotAI Kimi K2.5 | Adjusted LLM | 7 | 0.814 | 10 | 1.76 | Top 8 by COMET-Kiwi |
| Anthropic Claude Fable 5 | Adjusted LLM | 8 | 0.814 | 8 | 1.73 | Top 8 by both metrics |
| xAI Grok 4.20 (reasoning) | Adjusted LLM | 9 | 0.813 | 13 | 1.79 | Not shortlisted |
| Anthropic Claude Sonnet 5 | Adjusted LLM | 10 | 0.812 | 12 | 1.78 | Not shortlisted |
| OpenAI GPT-4.1 | Adjusted LLM | 11 | 0.810 | 4 | 1.67 | Top 8 by MetricX |
| Anthropic Claude Opus 4.8 | Adjusted LLM | 12 | 0.810 | 9 | 1.74 | Not shortlisted |
| OpenAI GPT-5.5 | Adjusted LLM | 13 | 0.810 | 15 | 1.87 | Not shortlisted |
| Translated Lara 2 | Adjusted NMT | 14 | 0.809 | 11 | 1.77 | Real-time place |
| Cohere Command A Translate | Adjusted LLM | 15 | 0.809 | 16 | 1.87 | Not shortlisted |
| DeepSeek V3.2 | Adjusted LLM | 16 | 0.807 | 17 | 1.96 | Not shortlisted |
| Google Translation LLM | Adjusted NMT | 17 | 0.806 | 25 | 2.09 | Real-time place |
| Google Gemma 4 | Adjusted LLM | 18 | 0.806 | 21 | 2.02 | Not shortlisted |
| Anthropic Claude Haiku 4.5 | Adjusted LLM | 19 | 0.804 | 24 | 2.08 | Not shortlisted |
| Tilde TildeOpen LLM engine | Adjusted NMT | 20 | 0.804 | 23 | 2.05 | Not shortlisted |
| Mistral Large 3 | Adjusted LLM | 21 | 0.803 | 14 | 1.80 | Not shortlisted |
| Microsoft Translator | Off-the-shelf NMT | 22 | 0.797 | 22 | 2.02 | Not shortlisted |
| OpenAI gpt-oss-120b | Adjusted LLM | 23 | 0.797 | 18 | 2.01 | Not shortlisted |
| Mistral Medium 3.1 | Adjusted LLM | 24 | 0.795 | 20 | 2.01 | Not shortlisted |
| MiniMax M2.5 | Adjusted LLM | 25 | 0.793 | 27 | 2.27 | Not shortlisted |
| ModernMT Enterprise Edition | Off-the-shelf NMT | 26 | 0.792 | 19 | 2.01 | Not shortlisted |
| Supertext LLM | Off-the-shelf NMT | 27 | 0.791 | 26 | 2.17 | Not shortlisted |
| Swiss AI Apertus-8B-2509 | Off-the-shelf LLM | 28 | 0.785 | 30 | 2.62 | Not shortlisted |
| Google Cloud Advanced Translation API | Off-the-shelf NMT | 29 | 0.784 | 31 | 2.68 | Not shortlisted |
| SYSTRAN PNMT API | Off-the-shelf NMT | 30 | 0.765 | 32 | 2.74 | Not shortlisted |
| AI21 Jamba Large 1.7 | Adjusted LLM | 31 | 0.761 | 28 | 2.55 | Not shortlisted |
| Alibaba Qwen3-235B-A22B | Adjusted LLM | 32 | 0.760 | 29 | 2.61 | Not shortlisted |
| Yandex Cloud Translate v2 | Off-the-shelf NMT | 33 | 0.756 | 34 | 2.96 | Not shortlisted |
| Cohere Tiny Aya | Off-the-shelf LLM | 34 | 0.756 | 33 | 2.84 | Not shortlisted |
| Amazon Translate | Off-the-shelf NMT | 35 | 0.752 | 35 | 3.08 | Not shortlisted |
| Alibaba Machine Translation API | Off-the-shelf NMT | 36 | 0.663 | 36 | 5.40 | Not shortlisted |
| AppTek NMT | Off-the-shelf NMT | 37 | 0.603 | 37 | 7.36 | Not shortlisted |
| OpenAI GPT-5.6 Luna | Adjusted LLM | — | 0.812 | — | 1.66 | Same model family as OpenAI GPT-5.6 Terra |
| OpenAI GPT-5.6 Sol | Adjusted LLM | — | 0.818 | — | 1.68 | Same model family as OpenAI GPT-5.6 Terra |
| xAI Grok 4.20 (non-reasoning) | Adjusted LLM | — | 0.807 | — | 1.77 | 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 37 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.
The solution on DeepL Next Gen, first on COMET-Kiwi, has 29.0 weighted errors per 1,000 words by Intento LQA and meets neither general translation quality nor tag handling. The one-step solution on Anthropic Claude Fable 5, eighth on both metrics, has the second fewest weighted errors on the shortlist, 12.9 per 1,000 words. COMET-Kiwi alone brought in MoonshotAI Kimi K2.5, and MetricX alone OpenAI GPT-4.1. Nearest the cut are xAI Grok 4.20 (reasoning), ninth on COMET-Kiwi, and Anthropic Claude Opus 4.8, ninth on MetricX.
| Quality level | Cheapest solution that meets every requirement of the level | Cost per million characters |
|---|---|---|
| Basic | Google Gemini 3.5 Flash, one-step LLM adjusted for Standard requirements | $3.46 |
| Standard | Google Gemini 3.1 Pro Preview, one-step LLM adjusted for Standard requirements | $4.66 |
| Advanced | Google Gemini 3.1 Pro Preview, two-step with the gender-neutrality agent | $4.90 |
| Model | US dollars per million characters | Class |
|---|---|---|
| MoonshotAI Kimi K2.5 | $1.37 | LLMs |
| z.ai GLM-5 | $2.08 | LLMs |
| OpenAI GPT-4.1 | $3.18 | LLMs |
| Google Gemini 3.5 Flash | $3.46 | LLMs |
| OpenAI GPT-5.6 Terra | $4.59 | LLMs |
| Google Gemini 3.1 Pro Preview | $4.66 | LLMs |
| Anthropic Claude Sonnet 4.6 | $6.98 | LLMs |
| Google Translation LLM | $20.62 | 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 | $29.37 | LLMs |
| Model | Tokens per character | Class |
|---|---|---|
| Anthropic Claude Opus 4.8 | 0.494 | LLMs |
| Anthropic Claude Sonnet 5 | 0.492 | LLMs |
| Anthropic Claude Fable 5 | 0.491 | LLMs |
| AI21 Jamba Large 1.7 | 0.453 | LLMs |
| Mistral Medium 3.1 | 0.445 | LLMs |
| Mistral Large 3 | 0.442 | LLMs |
| Alibaba Qwen3-235B-A22B | 0.408 | Open-weight LLMs |
| Anthropic Claude Sonnet 4.6 | 0.397 | LLMs |
| Anthropic Claude Haiku 4.5 | 0.395 | LLMs |
| MoonshotAI Kimi K2.5 | 0.393 | LLMs |
| DeepSeek V3.2 | 0.379 | LLMs |
| z.ai GLM-5 | 0.378 | LLMs |
| Cohere Command A Translate | 0.352 | LLMs |
| OpenAI GPT-4.1 | 0.331 | LLMs |
| OpenAI GPT-5.5 | 0.331 | LLMs |
| OpenAI GPT-5.6 Luna | 0.331 | LLMs |
| OpenAI GPT-5.6 Terra | 0.331 | LLMs |
| OpenAI GPT-5.6 Sol | 0.329 | LLMs |
| Google Gemini 3.1 Pro Preview | 0.328 | LLMs |
| Google Gemini 3.5 Flash | 0.328 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 0.327 | LLMs |
| xAI Grok 4.20 (reasoning) | 0.327 | LLMs |
| MiniMax M2.5 | 0.327 | LLMs |
| Google Gemma 4 | 0.326 | Open-weight LLMs |
| Cohere Tiny Aya | 0.296 | LLMs |
| 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 | 29 |
| Standard | Anthropic Claude Fable 5, one-step LLM adjusted for Standard requirements | 66 |
| Advanced | Google Gemini 3.1 Pro Preview, two-step with the gender-neutrality agent | 108 |
| Model | Seconds per 1,000 words | Class |
|---|---|---|
| Microsoft Translator | 0.2 | MT engines |
| Google Cloud Advanced Translation API | 0.2 | MT engines |
| Google Translation LLM | 0.4 | LLM-based MT engines |
| DeepL Next Gen | 0.6 | LLM-based MT engines |
| ModernMT Enterprise Edition | 0.6 | MT engines |
| Translated Lara 2 | 1.4 | LLM-based MT engines |
| Amazon Translate | 1.4 | MT engines |
| SYSTRAN PNMT API | 2.2 | MT engines |
| OpenAI gpt-oss-120b | 2.8 | Open-weight LLMs |
| Yandex Cloud Translate v2 | 3.8 | MT engines |
| Tilde TildeOpen LLM engine | 9.2 | LLM-based MT engines |
| OpenAI GPT-5.6 Luna | 14 | LLMs |
| Supertext LLM | 16 | LLM-based MT engines |
| Cohere Tiny Aya | 19 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 20 | LLMs |
| OpenAI GPT-5.6 Terra | 29 | LLMs |
| OpenAI GPT-4.1 | 30 | LLMs |
| Anthropic Claude Haiku 4.5 | 35 | LLMs |
| MiniMax M2.5 | 42 | LLMs |
| Swiss AI Apertus-8B-2509 | 42 | Open-weight LLMs |
| Mistral Medium 3.1 | 45 | LLMs |
| Cohere Command A Translate | 47 | LLMs |
| Alibaba Machine Translation API | 50 | MT engines |
| xAI Grok 4.20 (reasoning) | 54 | LLMs |
| Google Gemini 3.5 Flash | 58 | LLMs |
| Anthropic Claude Sonnet 5 | 61 | LLMs |
| Mistral Large 3 | 61 | LLMs |
| OpenAI GPT-5.6 Sol | 61 | LLMs |
| DeepSeek V3.2 | 64 | LLMs |
| AI21 Jamba Large 1.7 | 65 | LLMs |
| Anthropic Claude Fable 5 | 66 | LLMs |
| OpenAI GPT-5.5 | 66 | LLMs |
| Anthropic Claude Sonnet 4.6 | 67 | LLMs |
| Anthropic Claude Opus 4.8 | 79 | LLMs |
| AppTek NMT | 83 | MT engines |
| Google Gemini 3.1 Pro Preview | 87 | LLMs |
| Alibaba Qwen3-235B-A22B | 153 | Open-weight LLMs |
| MoonshotAI Kimi K2.5 | 160 | LLMs |
| Google Gemma 4 | 170 | Open-weight LLMs |
| z.ai GLM-5 | 221 | 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 Slovak.
Nine models are real-time in Slovak: five MT engines, three LLM-based MT engines and the open-weight OpenAI gpt-oss-120b. Translations from the three real-time solutions on the shortlist have 5 to 8 major errors, and those from Google Translation LLM also have 9 critical errors. The fastest solution takes 29 seconds per 1,000 words under the Basic requirements, 66 under the Standard ones and 108 under the Advanced ones, as a two-step solution (translation + post-edit) with the gender-neutrality agent.
| 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 |
|---|---|---|---|---|---|
| Anthropic Claude Fable 5 | three-step, with the gender-neutrality agent | 100% | ✓ | $1,027 | 220 |
| Google Gemini 3.1 Pro Preview | two-step, with the gender-neutrality agent | 100% | ✓ | $4.90 | 108 |
With the gender-neutrality agent, the three-step solution (translation + LQA + repair) on Anthropic Claude Fable 5 has 5.3 weighted errors per 1,000 words, against 12.9 in one step, and the two-step solution (translation + post-edit) on Google Gemini 3.1 Pro Preview 17.8, against 15.5.
Even at its lower bound, the agreement of each pair of metrics is above its threshold: 0.63 against 0.53 for the quality-estimation pair, and 0.48 against 0.46 for the similarity pair. Neither agreement is borderline. The three LQA judges agree on a median of 96.6% of the sentences of the shortlisted solutions. So the results on this page rest on measurement we can rely on.
Leave feedback