The State of Translation Automation 2026 · Languages
What it takes to automate translation from English into Kyrgyz 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 Kyrgyz. |
| Tone of voice | The formal register (Сиз) 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 Kyrgyz, 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. 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 three-step solutions (translation + LQA + repair) meet every Standard requirement in Kyrgyz, and neither has the fewest errors, lowest cost and shortest time together. The one on Anthropic Claude Fable 5 has 8.9 total errors per 1,000 words, against 11.2 for the one on OpenAI GPT-5.6 Terra. The one on OpenAI GPT-5.6 Terra costs $553 against $623 per million characters and takes 196 against 314 seconds per 1,000 words. Optimizing for quality selects Anthropic Claude Fable 5, and for cost or latency OpenAI GPT-5.6 Terra.
A cost cap of $552 per million characters or less in Kyrgyz leaves only one-step solutions, at $1.48 to $33.98, and two-step solutions (translation + post-edit), at $34.99 and $75.40. None of those with a conclusive Intento LQA result meet even the Basic requirements. Three of the one-step solutions have no such result. The three-step solutions that meet every Basic requirement take 196 to 407 seconds per 1,000 words, the one on Google Gemini 3.1 Pro Preview the longest.
The Advanced requirements add gender neutrality to the Standard ones. Both three-step solutions that meet every Standard requirement in Kyrgyz meet it too, without the gender-neutrality agent. So meeting the Advanced requirements costs the same and takes the same time as meeting the Standard ones, at least $552 per million characters and 195 seconds per 1,000 words, with the same total errors.
Anthropic Claude Fable 5
three-step
OpenAI GPT-5.6 Terra
three-step
OpenAI GPT-5.6 Terra
three-step
Anthropic Claude Fable 5
three-step
OpenAI GPT-5.6 Terra
three-step
OpenAI GPT-5.6 Terra
three-step
Anthropic Claude Fable 5
three-step
OpenAI GPT-5.6 Terra
three-step
OpenAI GPT-5.6 Terra
three-step
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 | ☆☆☆ | three-step | OpenAI GPT-5.6 Terra | 11.2 | $553 | 196 |
| Standard | Tier 2: two or more solutions, none simple | ☆☆ | three-step | OpenAI GPT-5.6 Terra | 11.2 | $553 | 196 |
| Advanced | Tier 2: two or more solutions, none simple | ☆☆ | three-step | OpenAI GPT-5.6 Terra | 11.2 | $553 | 196 |
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.
Three three-step solutions (translation + LQA + repair) meet every Basic requirement in Kyrgyz. The Standard requirements remove the one on Google Gemini 3.1 Pro Preview, on formatting, which the one-step solution on the same model meets. The Advanced requirements remove neither of the other two. The simplest solution is three-step for all three sets of requirements, and the gender-neutrality agent is not needed. Optimizing for quality selects Anthropic Claude Fable 5 each time, and optimizing for cost or latency OpenAI GPT-5.6 Terra.
| Model | US dollars per million characters | Weighted errors per 1,000 words | Group |
|---|---|---|---|
| xAI Grok 4.20 (non-reasoning) | 1.48 | 79.4 | Misses a Standard requirement |
| Google Gemini 3.5 Flash | 4.2 | 47.5 | Misses a Standard requirement |
| OpenAI GPT-5.6 Terra | 4.82 | 25.7 | Misses a Standard requirement |
| Google Gemini 3.1 Pro Preview | 5.53 | 38.9 | Misses a Standard requirement |
| OpenAI GPT-5.5 | 12.04 | 20.4 | Misses a Standard requirement |
| Anthropic Claude Opus 4.8 | 16.74 | 43.8 | Misses a Standard requirement |
| Google Translation LLM | 21.39 | 77.1 | Misses a Standard requirement |
| Anthropic Claude Fable 5 | 33.98 | 28.3 | Misses a Standard requirement |
Among the one-step solutions in Kyrgyz, a higher price does not always come with fewer errors. The dearest, on Anthropic Claude Fable 5 at $33.98 per million characters, has 28.3 weighted errors per 1,000 words, more than those on OpenAI GPT-5.5 at $12.04 with 20.4 and OpenAI GPT-5.6 Terra at $4.82 with 25.7. The one on Google Translation LLM costs $21.39 per million characters and has 77.1 weighted errors per 1,000 words, close to the 79.4 of the one on xAI Grok 4.20 (non-reasoning) at $1.48.
| Requirement | Median errors per 1,000 words |
|---|---|
| General translation quality | 41.3 |
| Tag handling | 0.7 |
| Tone of voice | 0 |
| Gender neutrality | 0 |
The requirements with countable errors. Every solution with its result on each requirement: the explorer.
The 8 shortlisted one-step solutions with a conclusive Intento LQA result have 20.4 to 79.4 weighted errors per 1,000 words in Kyrgyz, over the bar of general translation quality. Two-step solutions (translation + post-edit) on two of these models have 26.7 and 33.6 weighted errors per 1,000 words. Three-step solutions (translation + LQA + repair) bring three models under the bar, with 8.9 weighted errors per 1,000 words on Anthropic Claude Fable 5, 11.2 on OpenAI GPT-5.6 Terra and 18.1 on Google Gemini 3.1 Pro Preview.
| Model | Weighted errors per 1,000 words of the shortlisted one-step solutions in Kyrgyz | Class |
|---|---|---|
| xAI Grok 4.20 (non-reasoning) | 79.4 | LLMs |
| Google Translation LLM | 77.1 | LLM-based MT engines |
| Google Gemini 3.5 Flash | 47.5 | LLMs |
| Anthropic Claude Opus 4.8 | 43.8 | LLMs |
| Google Gemini 3.1 Pro Preview | 38.9 | LLMs |
| Anthropic Claude Fable 5 | 28.3 | LLMs |
| OpenAI GPT-5.6 Terra | 25.7 | LLMs |
| OpenAI GPT-5.5 | 20.4 | LLMs |
No translation into Kyrgyz: AI21 Jamba Large 1.7 (garbage or near-empty).
In one step, translations from six of the seven shortlisted LLMs with a conclusive Intento LQA result have 20.4 to 47.5 weighted errors per 1,000 words in Kyrgyz. Translations from the seventh, xAI Grok 4.20 (non-reasoning), have 79.4 weighted errors per 1,000 words, over the limit of 60 in our language-support criteria. Translations from Google Translation LLM, the one LLM-based MT engine with a conclusive result, have 7 critical errors. None of the MT engines that are not LLM-based made the shortlist.
| Model | COMET-Kiwi | Rank on COMET-Kiwi | MetricX | Rank on MetricX | Shortlisted | Shortlist place |
|---|---|---|---|---|---|---|
| Anthropic Claude Sonnet 5 | 0.781 | 11 | 2.10 | 13 | No | Not shortlisted |
| Anthropic Claude Sonnet 4.6 | 0.780 | 12 | 1.95 | 10 | No | Not shortlisted |
| OpenAI GPT-4.1 | 0.779 | 13 | 2.00 | 11 | No | Not shortlisted |
| DeepSeek V3.2 | 0.777 | 15 | 2.06 | 12 | No | Not shortlisted |
| MoonshotAI Kimi K2.5 | 0.775 | 16 | 2.16 | 16 | No | Not shortlisted |
| Google Gemma 4 | 0.774 | 17 | 2.17 | 17 | No | Not shortlisted |
| Microsoft Translator | 0.774 | 18 | 2.15 | 15 | No | Not shortlisted |
| Mistral Large 3 | 0.762 | 19 | 2.22 | 18 | No | Not shortlisted |
| OpenAI gpt-oss-120b | 0.759 | 20 | 2.64 | 21 | No | Not shortlisted |
| Google Cloud Advanced Translation API | 0.756 | 21 | 2.79 | 22 | No | Not shortlisted |
| ModernMT Enterprise Edition | 0.755 | 22 | 2.42 | 20 | No | Not shortlisted |
| NiuTrans | 0.751 | 23 | 3.01 | 25 | No | Not shortlisted |
| Mistral Medium 3.1 | 0.746 | 24 | 2.82 | 23 | No | Not shortlisted |
| Anthropic Claude Haiku 4.5 | 0.743 | 25 | 3.24 | 26 | No | Not shortlisted |
| Swiss AI Apertus-8B-2509 | 0.741 | 26 | 2.90 | 24 | No | Not shortlisted |
| Yandex Cloud Translate v2 | 0.737 | 27 | 3.35 | 28 | No | Not shortlisted |
| MiniMax M2.5 | 0.731 | 28 | 3.33 | 27 | No | Not shortlisted |
| Cohere Command A Translate | 0.656 | 29 | 5.46 | 29 | No | Not shortlisted |
| OpenAI GPT-5.6 Terra | 0.797 | 1 | 1.79 | 6 | Yes | Top 8 by both metrics |
| OpenAI GPT-5.5 | 0.793 | 2 | 1.77 | 5 | Yes | Top 8 by both metrics |
| Google Gemini 3.1 Pro Preview | 0.791 | 3 | 1.60 | 2 | Yes | Top 8 by both metrics |
| DeepL Next Gen | 0.790 | 4 | 1.91 | 9 | Yes | Top 8 by COMET-Kiwi |
| xAI Grok 4.20 (non-reasoning) | 0.789 | 5 | 1.75 | 4 | Yes | Top 8 by both metrics |
| Google Gemini 3.5 Flash | 0.788 | 6 | 1.56 | 1 | Yes | Top 8 by both metrics |
| Anthropic Claude Fable 5 | 0.787 | 7 | 1.68 | 3 | Yes | Top 8 by both metrics |
| Translated Lara 2 | 0.786 | 8 | 2.11 | 14 | Yes | Top 8 by COMET-Kiwi |
| Anthropic Claude Opus 4.8 | 0.784 | 9 | 1.90 | 8 | Yes | Top 8 by MetricX |
| z.ai GLM-5 | 0.781 | 10 | 1.84 | 7 | Yes | Top 8 by MetricX |
| Google Translation LLM | 0.778 | 14 | 2.27 | 19 | Yes | Real-time place |
| Model | Solution complexity | COMET-Kiwi rank | COMET-Kiwi | MetricX rank | MetricX | Shortlist |
|---|---|---|---|---|---|---|
| OpenAI GPT-5.6 Terra | Adjusted LLM | 1 | 0.797 | 6 | 1.79 | Top 8 by both metrics |
| OpenAI GPT-5.5 | Adjusted LLM | 2 | 0.793 | 5 | 1.77 | Top 8 by both metrics |
| Google Gemini 3.1 Pro Preview | Adjusted LLM | 3 | 0.791 | 2 | 1.60 | Top 8 by both metrics |
| DeepL Next Gen | Off-the-shelf NMT | 4 | 0.790 | 9 | 1.91 | Top 8 by COMET-Kiwi |
| xAI Grok 4.20 (non-reasoning) | Adjusted LLM | 5 | 0.789 | 4 | 1.75 | Top 8 by both metrics |
| Google Gemini 3.5 Flash | Adjusted LLM | 6 | 0.788 | 1 | 1.56 | Top 8 by both metrics |
| Anthropic Claude Fable 5 | Adjusted LLM | 7 | 0.787 | 3 | 1.68 | Top 8 by both metrics |
| Translated Lara 2 | Adjusted NMT | 8 | 0.786 | 14 | 2.11 | Top 8 by COMET-Kiwi |
| Anthropic Claude Opus 4.8 | Adjusted LLM | 9 | 0.784 | 8 | 1.90 | Top 8 by MetricX |
| z.ai GLM-5 | Adjusted LLM | 10 | 0.781 | 7 | 1.84 | Top 8 by MetricX |
| Anthropic Claude Sonnet 5 | Adjusted LLM | 11 | 0.781 | 13 | 2.10 | Not shortlisted |
| Anthropic Claude Sonnet 4.6 | Adjusted LLM | 12 | 0.780 | 10 | 1.95 | Not shortlisted |
| OpenAI GPT-4.1 | Adjusted LLM | 13 | 0.779 | 11 | 2.00 | Not shortlisted |
| Google Translation LLM | Adjusted NMT | 14 | 0.778 | 19 | 2.27 | Real-time place |
| DeepSeek V3.2 | Adjusted LLM | 15 | 0.777 | 12 | 2.06 | Not shortlisted |
| MoonshotAI Kimi K2.5 | Adjusted LLM | 16 | 0.775 | 16 | 2.16 | Not shortlisted |
| Google Gemma 4 | Adjusted LLM | 17 | 0.774 | 17 | 2.17 | Not shortlisted |
| Microsoft Translator | Off-the-shelf NMT | 18 | 0.774 | 15 | 2.15 | Not shortlisted |
| Mistral Large 3 | Adjusted LLM | 19 | 0.762 | 18 | 2.22 | Not shortlisted |
| OpenAI gpt-oss-120b | Adjusted LLM | 20 | 0.759 | 21 | 2.64 | Not shortlisted |
| Google Cloud Advanced Translation API | Off-the-shelf NMT | 21 | 0.756 | 22 | 2.79 | Not shortlisted |
| ModernMT Enterprise Edition | Off-the-shelf NMT | 22 | 0.755 | 20 | 2.42 | Not shortlisted |
| NiuTrans | Off-the-shelf NMT | 23 | 0.751 | 25 | 3.01 | Not shortlisted |
| Mistral Medium 3.1 | Adjusted LLM | 24 | 0.746 | 23 | 2.82 | Not shortlisted |
| Anthropic Claude Haiku 4.5 | Adjusted LLM | 25 | 0.743 | 26 | 3.24 | Not shortlisted |
| Swiss AI Apertus-8B-2509 | Off-the-shelf LLM | 26 | 0.741 | 24 | 2.90 | Not shortlisted |
| Yandex Cloud Translate v2 | Off-the-shelf NMT | 27 | 0.737 | 28 | 3.35 | Not shortlisted |
| MiniMax M2.5 | Adjusted LLM | 28 | 0.731 | 27 | 3.33 | Not shortlisted |
| Cohere Command A Translate | Adjusted LLM | 29 | 0.656 | 29 | 5.46 | Not shortlisted |
| OpenAI GPT-5.6 Luna | Adjusted LLM | — | 0.793 | — | 1.81 | Same model family as OpenAI GPT-5.6 Terra |
| OpenAI GPT-5.6 Sol | Adjusted LLM | — | 0.791 | — | 1.73 | Same model family as OpenAI GPT-5.6 Terra |
| xAI Grok 4.20 (reasoning) | Adjusted LLM | — | 0.785 | — | 1.91 | Same 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 29 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.
We shortlisted DeepL Next Gen and Translated Lara 2 on COMET-Kiwi alone, at ranks 4 and 8. MetricX ranks them 9 and 14. Neither has a conclusive Intento LQA result. We shortlisted Anthropic Claude Opus 4.8 and z.ai GLM-5 on MetricX alone. Of the one-step solutions with a conclusive Intento LQA result, the one on OpenAI GPT-5.5 has the fewest weighted errors. COMET-Kiwi ranks the model second and MetricX fifth. Nearest the cut are Anthropic Claude Sonnet 5, rank 11 on COMET-Kiwi, and Anthropic Claude Sonnet 4.6, rank 10 on MetricX.
| Quality level | Cheapest solution that meets every requirement of the level | Cost per million characters |
|---|---|---|
| Basic | OpenAI GPT-5.6 Terra, three-step | $553 |
| Standard | OpenAI GPT-5.6 Terra, three-step | $553 |
| Advanced | OpenAI GPT-5.6 Terra, three-step | $553 |
| Model | US dollars per million characters | Class |
|---|---|---|
| xAI Grok 4.20 (non-reasoning) | $1.48 | LLMs |
| z.ai GLM-5 | $2.68 | LLMs |
| Google Gemini 3.5 Flash | $4.2 | LLMs |
| OpenAI GPT-5.6 Terra | $4.82 | LLMs |
| Google Gemini 3.1 Pro Preview | $5.53 | LLMs |
| OpenAI GPT-5.5 | $12.04 | LLMs |
| Anthropic Claude Opus 4.8 | $16.74 | LLMs |
| Google Translation LLM | $21.39 | 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 | $33.98 | LLMs |
| Model | Tokens per character | Class |
|---|---|---|
| Mistral Large 3 | 0.547 | LLMs |
| Anthropic Claude Fable 5 | 0.544 | LLMs |
| Anthropic Claude Sonnet 5 | 0.544 | LLMs |
| Anthropic Claude Opus 4.8 | 0.543 | LLMs |
| Mistral Medium 3.1 | 0.543 | LLMs |
| Anthropic Claude Sonnet 4.6 | 0.515 | LLMs |
| Anthropic Claude Haiku 4.5 | 0.512 | LLMs |
| DeepSeek V3.2 | 0.482 | LLMs |
| z.ai GLM-5 | 0.481 | LLMs |
| Cohere Command A Translate | 0.455 | LLMs |
| MiniMax M2.5 | 0.452 | LLMs |
| MoonshotAI Kimi K2.5 | 0.443 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 0.428 | LLMs |
| xAI Grok 4.20 (reasoning) | 0.427 | LLMs |
| Google Gemini 3.5 Flash | 0.37 | LLMs |
| Google Gemini 3.1 Pro Preview | 0.369 | LLMs |
| Google Gemma 4 | 0.369 | Open-weight LLMs |
| OpenAI GPT-5.6 Terra | 0.326 | LLMs |
| OpenAI GPT-5.6 Sol | 0.324 | LLMs |
| OpenAI GPT-4.1 | 0.322 | LLMs |
| OpenAI GPT-5.5 | 0.322 | LLMs |
| OpenAI GPT-5.6 Luna | 0.322 | LLMs |
The tokenizers of the Anthropic Claude models produce 0.512 to 0.544 tokens per character of the Kyrgyz translation, the OpenAI ones 0.322 to 0.326. Those of Anthropic Claude Fable 5 and Anthropic Claude Opus 4.8 produce almost the same, 0.544 and 0.543 tokens per character, yet the one-step solutions on these models cost $33.98 and $16.74 per million characters. The steps raise the cost. The solution on OpenAI GPT-5.6 Terra costs $4.82 per million characters in one step and $553 in three steps (translation + LQA + repair).
| Quality level | Fastest solution that meets every requirement of the level | Seconds per 1,000 words |
|---|---|---|
| Basic | OpenAI GPT-5.6 Terra, three-step | 196 |
| Standard | OpenAI GPT-5.6 Terra, three-step | 196 |
| Advanced | OpenAI GPT-5.6 Terra, three-step | 196 |
| Model | Seconds per 1,000 words | Class |
|---|---|---|
| Microsoft Translator | 0.4 | MT engines |
| Google Cloud Advanced Translation API | 0.4 | MT engines |
| ModernMT Enterprise Edition | 0.6 | MT engines |
| DeepL Next Gen | 1.2 | LLM-based MT engines |
| Google Translation LLM | 1.5 | LLM-based MT engines |
| Translated Lara 2 | 2.1 | LLM-based MT engines |
| OpenAI gpt-oss-120b | 2.5 | Open-weight LLMs |
| Yandex Cloud Translate v2 | 5.2 | MT engines |
| OpenAI GPT-5.6 Luna | 17 | LLMs |
| Google Gemini 3.5 Flash | 22 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 30 | LLMs |
| OpenAI GPT-4.1 | 37 | LLMs |
| OpenAI GPT-5.6 Terra | 51 | LLMs |
| Swiss AI Apertus-8B-2509 | 55 | Open-weight LLMs |
| NiuTrans | 58 | MT engines |
| Mistral Medium 3.1 | 61 | LLMs |
| Anthropic Claude Haiku 4.5 | 62 | LLMs |
| MiniMax M2.5 | 68 | LLMs |
| xAI Grok 4.20 (reasoning) | 82 | LLMs |
| Cohere Command A Translate | 88 | LLMs |
| DeepSeek V3.2 | 95 | LLMs |
| Google Gemini 3.1 Pro Preview | 97 | LLMs |
| Mistral Large 3 | 100 | LLMs |
| OpenAI GPT-5.5 | 104 | LLMs |
| Anthropic Claude Sonnet 5 | 116 | LLMs |
| Anthropic Claude Sonnet 4.6 | 117 | LLMs |
| Anthropic Claude Fable 5 | 125 | LLMs |
| OpenAI GPT-5.6 Sol | 125 | LLMs |
| Anthropic Claude Opus 4.8 | 148 | LLMs |
| Google Gemma 4 | 180 | Open-weight LLMs |
| z.ai GLM-5 | 311 | LLMs |
| MoonshotAI Kimi K2.5 | 334 | 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 Kyrgyz.
The seven real-time models in Kyrgyz are three MT engines, three LLM-based MT engines and the open-weight OpenAI gpt-oss-120b. Three of them are on the shortlist: DeepL Next Gen, Translated Lara 2 and Google Translation LLM. Translations from Google Translation LLM have 77.1 weighted errors per 1,000 words, and the other two have no conclusive Intento LQA result. None of the three meet tag handling. Among the LLMs, translations from OpenAI gpt-oss-120b and OpenAI GPT-5.6 Luna are the fastest, at 2.5 and 17 seconds per 1,000 words.
| 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 | 100% | ✓ | $623 | 314 |
| OpenAI GPT-5.6 Terra | three-step | 100% | ✓ | $553 | 196 |
For Kyrgyz we check the formal register, with Сиз. Translations from all 11 shortlisted one-step solutions meet tone of voice, including the three without a conclusive Intento LQA result. Kyrgyz has no grammatical gender for persons, so a translation is neutral by default. We check that it avoids gendered terms and does not assume a gender the source leaves open. In translations from both three-step solutions (translation + LQA + repair) that meet every Standard requirement, 100% of the sentences about people are gender-neutral.
The three LQA judges agree on a median of 93.2% of the sentences across the 11 shortlisted solutions in Kyrgyz. With the LQA judges in agreement at this level, and both pairs of automatic metrics in agreement too, the Intento LQA error counts and the shortlist on this page carry weight.
Leave feedback