The State of Translation Automation 2026 · Languages
What it takes to automate translation from English into Turkmen 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 Turkmen. |
| Tone of voice | The formal register (Siz) 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. |
| Metric pair | Metrics | Agreement in Turkmen | 90% band | Minimum | Result |
|---|---|---|---|---|---|
| Similarity | LaBSE, Cohere embed v3 | 0.362 | 0.343 to 0.378 | 0.461 | Under the minimum |
| Quality estimation | COMET-Kiwi, MetricX | — | — | 0.534 | Cannot be applied |
Pick a quality level, or your own requirements, bars and constraints, to see the shortlisted solutions that meet them in Turkmen, with a column for each requirement you select. Under Optimize for, choose which of them comes first: the one with the lowest cost or the shortest time. Click a column heading to sort by it.
We did not shortlist solutions in Turkmen, so the explorer has none: why.
We have a price for 4 of the 6 real-time (<2 sec) models: Microsoft Translator costs $10.00 per million characters, Google Cloud Advanced Translation API $20.00, Translated Lara 2 $24.99 and DeepL Next Gen $27.50. Two of the three LLMs added by a limit of 33 seconds per 1,000 words cost less than any priced real-time model: xAI Grok 4.20 (non-reasoning) at $1.28 per million characters and OpenAI GPT-4.1 at $3.09. They take 29 and 33 seconds per 1,000 words. OpenAI GPT-5.6 Luna costs $0.46 per million characters, a lower bound as it has reasoning.
We cannot assess general translation quality in Turkmen: how confident we are.
We cannot assess general translation quality in Turkmen: how confident we are.
We cannot assess general translation quality in Turkmen: how confident we are.
No translation into Turkmen: AI21 Jamba Large 1.7 (garbage or near-empty, wrong language), Swiss AI Apertus-8B-2509 (wrong language, garbage or near-empty).
Two vendors have models with and without support for Turkmen. Google Translation LLM has no support for it, and we got translations from Google Cloud Advanced Translation API and three other Google models. Cohere Tiny Aya has no support for it, and we got translations from Cohere Command A Translate. None of the three Widn AI models has support for Turkmen, and neither of the two Alibaba models has.
What stays on this page does not rest on a ranking: which models we got translations into Turkmen from, the off-the-shelf cost of each model with a price, the time of each model we timed, the tokens per character, the length of a translation and the tone of voice and gender-neutrality requirements of Turkmen.
| Model | US dollars per million characters | Class |
|---|---|---|
| Google Gemma 4 | $0.17 | Open-weight LLMs |
| OpenAI GPT-5.6 Luna | $0.46 | LLMs |
| MiniMax M2.5 | $0.72 | LLMs |
| Mistral Large 3 | $0.74 | LLMs |
| Mistral Medium 3.1 | $1.06 | LLMs |
| DeepSeek V3.2 | $1.09 | LLMs |
| xAI Grok 4.20 (non-reasoning) | $1.28 | LLMs |
| xAI Grok 4.20 (reasoning) | $1.28 | LLMs |
| MoonshotAI Kimi K2.5 | $1.47 | LLMs |
| z.ai GLM-5 | $2.18 | LLMs |
| Anthropic Claude Haiku 4.5 | $2.56 | LLMs |
| OpenAI GPT-4.1 | $3.09 | LLMs |
| Google Gemini 3.5 Flash | $4.21 | LLMs |
| OpenAI GPT-5.6 Terra | $4.56 | LLMs |
| Google Gemini 3.1 Pro Preview | $5.63 | LLMs |
| Anthropic Claude Sonnet 5 | $6.75 | LLMs |
| OpenAI GPT-5.6 Sol | $7.96 | LLMs |
| Anthropic Claude Sonnet 4.6 | $8.1 | LLMs |
| Microsoft Translator | $10 | MT engines |
| OpenAI GPT-5.5 | $11.43 | LLMs |
| Anthropic Claude Opus 4.8 | $16.96 | LLMs |
| Google Cloud Advanced Translation API | $20 | MT engines |
| Translated Lara 2 | $24.99 | LLM-based MT engines |
| DeepL Next Gen | $27.5 | LLM-based MT engines |
| Anthropic Claude Fable 5 | $33.96 | LLMs |
| Model | Tokens per character | Class |
|---|---|---|
| Anthropic Claude Sonnet 5 | 0.563 | LLMs |
| Anthropic Claude Fable 5 | 0.56 | LLMs |
| Anthropic Claude Opus 4.8 | 0.559 | LLMs |
| Mistral Large 3 | 0.484 | LLMs |
| Mistral Medium 3.1 | 0.47 | LLMs |
| Anthropic Claude Sonnet 4.6 | 0.458 | LLMs |
| Anthropic Claude Haiku 4.5 | 0.446 | LLMs |
| DeepSeek V3.2 | 0.425 | LLMs |
| MoonshotAI Kimi K2.5 | 0.417 | LLMs |
| z.ai GLM-5 | 0.4 | LLMs |
| Google Gemini 3.1 Pro Preview | 0.39 | LLMs |
| Google Gemini 3.5 Flash | 0.388 | LLMs |
| Google Gemma 4 | 0.385 | Open-weight LLMs |
| xAI Grok 4.20 (non-reasoning) | 0.375 | LLMs |
| xAI Grok 4.20 (reasoning) | 0.373 | LLMs |
| Cohere Command A Translate | 0.353 | LLMs |
| MiniMax M2.5 | 0.353 | LLMs |
| OpenAI GPT-5.6 Sol | 0.314 | LLMs |
| OpenAI GPT-4.1 | 0.313 | LLMs |
| OpenAI GPT-5.6 Luna | 0.312 | LLMs |
| OpenAI GPT-5.5 | 0.311 | LLMs |
| OpenAI GPT-5.6 Terra | 0.311 | LLMs |
Translations from the five OpenAI models with a tokenizer result have 0.311 to 0.314 tokens per character. They cost $0.46 to $11.43 per million characters, lower bounds for the four with reasoning, so tokens per character alone do not set the cost. Translations from Mistral Large 3 have 0.484 tokens per character and cost $0.74 per million characters, less than those from Anthropic Claude Haiku 4.5, at 0.446 and $2.56.
| 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 |
| DeepL Next Gen | 0.6 | LLM-based MT engines |
| OpenAI gpt-oss-120b | 2.4 | Open-weight LLMs |
| Translated Lara 2 | 4.2 | LLM-based MT engines |
| OpenAI GPT-5.6 Luna | 18 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 29 | LLMs |
| OpenAI GPT-4.1 | 33 | LLMs |
| OpenAI GPT-5.6 Terra | 47 | LLMs |
| Anthropic Claude Haiku 4.5 | 52 | LLMs |
| Mistral Medium 3.1 | 54 | LLMs |
| Cohere Command A Translate | 63 | LLMs |
| MiniMax M2.5 | 69 | LLMs |
| Google Gemini 3.5 Flash | 72 | LLMs |
| DeepSeek V3.2 | 74 | LLMs |
| xAI Grok 4.20 (reasoning) | 81 | LLMs |
| Mistral Large 3 | 87 | LLMs |
| Anthropic Claude Sonnet 5 | 92 | LLMs |
| OpenAI GPT-5.5 | 94 | LLMs |
| Anthropic Claude Fable 5 | 100 | LLMs |
| Anthropic Claude Sonnet 4.6 | 100 | LLMs |
| Google Gemini 3.1 Pro Preview | 101 | LLMs |
| OpenAI GPT-5.6 Sol | 104 | LLMs |
| Anthropic Claude Opus 4.8 | 113 | LLMs |
| Google Gemma 4 | 218 | Open-weight LLMs |
| z.ai GLM-5 | 232 | LLMs |
| MoonshotAI Kimi K2.5 | 249 | 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 Turkmen.
Translations from both LLM-based MT engines, DeepL Next Gen and Translated Lara 2, count as real-time, and into Turkmen they take 0.6 and 4.2 seconds per 1,000 words. Of the two open-weight LLMs we got translations from, translations from OpenAI gpt-oss-120b take 2.4 seconds per 1,000 words and those from Google Gemma 4 take 218.
Leave feedback