The State of Translation Automation 2026 · Languages
What it takes to automate translation from English into Albanian 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 Albanian. |
| Tone of voice | The formal register (Ju) 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 Albanian, 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.
Of the two solutions that meet every Standard requirement, the two-step solution (translation + post-edit) on OpenAI GPT-5.6 Terra costs $11.46 per million characters, takes 69 seconds per 1,000 words and has 23.0 total errors per 1,000 words. The one-step solution on Anthropic Claude Fable 5 costs $31.75 per million characters, takes 85 seconds per 1,000 words and has 23.6 total errors but fewer weighted errors, 18.1 per 1,000 words against 18.5. Optimize for selects OpenAI GPT-5.6 Terra for quality, cost and latency.
None of the solutions meet every Standard requirement in Albanian under a cost cap of at most $11.45 per million characters or a time limit of at most 68 seconds per 1,000 words. A limit of one step leaves the solution on Anthropic Claude Fable 5 alone, at $31.75 per million characters and 85 seconds per 1,000 words. Under that cap, six one-step solutions on LLMs, from $0.16 to $6.29 per million characters, meet tone of voice, tag handling and formatting, but not general translation quality.
For the Advanced requirements, the gender-neutrality agent works with both solutions. With the two-step solution on OpenAI GPT-5.6 Terra, it raises the cost from $11.46 to $13.29 per million characters and the time from 69 to 84 seconds per 1,000 words. Total errors fall from 23.0 to 22.1 per 1,000 words, and weighted errors rise from 18.5 to 21.4 per 1,000 words, more than general translation quality allows. With the agent, the solution on Anthropic Claude Fable 5 costs $62.86 per million characters.
OpenAI GPT-5.6 Terra
two-step
OpenAI GPT-5.6 Terra
two-step
OpenAI GPT-5.6 Terra
two-step
OpenAI GPT-5.6 Terra
two-step
OpenAI GPT-5.6 Terra
two-step
OpenAI GPT-5.6 Terra
two-step
Anthropic Claude Fable 5
two-step with the gender-neutrality agent
Misses gender neutrality.
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 | Anthropic Claude Fable 5 | 23.6 | $31.75 | 85 |
| Standard | Tier 1: simple solution + backup | ★ ☆ | one-step | Anthropic Claude Fable 5 | 23.6 | $31.75 | 85 |
| Advanced | Tier 4: no solution | No solution | — | — | — |
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.
Both solutions that meet every Basic requirement also meet every Standard requirement: the one-step solution on Anthropic Claude Fable 5 and the two-step one (translation + post-edit) on OpenAI GPT-5.6 Terra. For both levels, Optimize for selects OpenAI GPT-5.6 Terra for quality, cost and latency. The Advanced requirements add gender neutrality, which neither meets, even with the gender-neutrality agent. With it, the solution on Anthropic Claude Fable 5 has 18.5 total errors per 1,000 words and the one on OpenAI GPT-5.6 Terra 22.1.
| Model | US dollars per million characters | Weighted errors per 1,000 words | Group |
|---|---|---|---|
| Google Gemma 4 | 0.16 | 51.7 | Misses a Standard requirement |
| xAI Grok 4.20 (reasoning) | 1.23 | 40.9 | Misses a Standard requirement |
| OpenAI GPT-4.1 | 3.31 | 29.3 | Misses a Standard requirement |
| Google Gemini 3.5 Flash | 3.94 | 21.8 | Misses a Standard requirement |
| OpenAI GPT-5.6 Terra | 4.84 | 21.1 | Misses a Standard requirement |
| Google Gemini 3.1 Pro Preview | 5.29 | 20.8 | Misses a Standard requirement |
| Anthropic Claude Sonnet 5 | 6.29 | 33.3 | Misses a Standard requirement |
| Microsoft Translator | 10 | 54.4 | Misses a Standard requirement |
| Translated Lara 2 | 24.99 | 34.9 | Misses a Standard requirement |
| DeepL Next Gen | 27.5 | 36.3 | Misses a Standard requirement |
| Anthropic Claude Fable 5 | 31.75 | 18.1 | Meets every Standard requirement |
| Requirement | Median errors per 1,000 words |
|---|---|
| General translation quality | 33.3 |
| Gender neutrality | 5.5 |
| Tone of voice | 0 |
| Tag handling | 0 |
The requirements with countable errors. Every solution with its result on each requirement: the explorer.
| Model | Weighted errors per 1,000 words of the shortlisted one-step solutions in Albanian | Class |
|---|---|---|
| Microsoft Translator | 54.4 | MT engines |
| Google Gemma 4 | 51.7 | Open-weight LLMs |
| xAI Grok 4.20 (reasoning) | 40.9 | LLMs |
| DeepL Next Gen | 36.3 | LLM-based MT engines |
| Translated Lara 2 | 34.9 | LLM-based MT engines |
| Anthropic Claude Sonnet 5 | 33.3 | LLMs |
| OpenAI GPT-4.1 | 29.3 | LLMs |
| Google Gemini 3.5 Flash | 21.8 | LLMs |
| OpenAI GPT-5.6 Terra | 21.1 | LLMs |
| Google Gemini 3.1 Pro Preview | 20.8 | LLMs |
| Anthropic Claude Fable 5 | 18.1 | LLMs |
Translations from six of the seven shortlisted LLMs that are not open-weight have 18.1 to 33.3 weighted errors per 1,000 words, fewer than those from Translated Lara 2 and DeepL Next Gen, the two shortlisted LLM-based MT engines, with 34.9 and 36.3. Translations from xAI Grok 4.20 (reasoning) have 40.9 weighted errors per 1,000 words, from the open-weight Google Gemma 4 51.7 and from the MT engine Microsoft Translator 54.4. All eight shortlisted LLMs support Albanian by our measured criteria.
| Model | COMET-Kiwi | Rank on COMET-Kiwi | MetricX | Rank on MetricX | Shortlisted | Shortlist place |
|---|---|---|---|---|---|---|
| OpenAI GPT-5.5 | 0.789 | 11 | 1.84 | 15 | No | Not shortlisted |
| z.ai GLM-5 | 0.788 | 12 | 1.69 | 14 | No | Not shortlisted |
| Anthropic Claude Sonnet 4.6 | 0.787 | 13 | 1.66 | 12 | No | Not shortlisted |
| DeepSeek V3.2 | 0.787 | 14 | 1.88 | 18 | No | Not shortlisted |
| Anthropic Claude Opus 4.8 | 0.786 | 15 | 1.67 | 13 | No | Not shortlisted |
| Google Translation LLM | 0.780 | 17 | 1.98 | 23 | No | Not shortlisted |
| MoonshotAI Kimi K2.5 | 0.779 | 18 | 1.96 | 21 | No | Not shortlisted |
| Tilde TildeOpen LLM engine | 0.778 | 19 | 1.87 | 17 | No | Not shortlisted |
| Mistral Large 3 | 0.776 | 20 | 1.95 | 20 | No | Not shortlisted |
| OpenAI gpt-oss-120b | 0.776 | 21 | 1.86 | 16 | No | Not shortlisted |
| ModernMT Enterprise Edition | 0.774 | 22 | 1.90 | 19 | No | Not shortlisted |
| Anthropic Claude Haiku 4.5 | 0.772 | 23 | 2.03 | 24 | No | Not shortlisted |
| Swiss AI Apertus-8B-2509 | 0.772 | 24 | 2.15 | 26 | No | Not shortlisted |
| SpeakLeash Bielik-11B-v3.0-Instruct | 0.768 | 25 | 2.29 | 29 | No | Not shortlisted |
| Cohere Command A Translate | 0.763 | 26 | 1.97 | 22 | No | Not shortlisted |
| MiniMax M2.5 | 0.761 | 27 | 2.13 | 25 | No | Not shortlisted |
| Alibaba Qwen3-235B-A22B | 0.761 | 28 | 2.19 | 27 | No | Not shortlisted |
| Yandex Cloud Translate v2 | 0.757 | 29 | 2.21 | 28 | No | Not shortlisted |
| Google Cloud Advanced Translation API | 0.754 | 30 | 2.57 | 32 | No | Not shortlisted |
| Mistral Medium 3.1 | 0.753 | 31 | 2.31 | 30 | No | Not shortlisted |
| Amazon Translate | 0.746 | 32 | 2.53 | 31 | No | Not shortlisted |
| Supertext LLM | 0.746 | 33 | 2.69 | 33 | No | Not shortlisted |
| SYSTRAN PNMT API | 0.678 | 34 | 3.74 | 35 | No | Not shortlisted |
| AI21 Jamba Large 1.7 | 0.667 | 35 | 3.42 | 34 | No | Not shortlisted |
| Alibaba Machine Translation API | 0.645 | 36 | 4.25 | 36 | No | Not shortlisted |
| OpenAI GPT-5.6 Terra | 0.797 | 1 | 1.62 | 5 | Yes | Top 8 by both metrics |
| DeepL Next Gen | 0.794 | 2 | 1.47 | 1 | Yes | Top 8 by both metrics |
| Google Gemma 4 | 0.794 | 3 | 1.61 | 4 | Yes | Top 8 by both metrics |
| xAI Grok 4.20 (reasoning) | 0.794 | 4 | 1.64 | 6 | Yes | Top 8 by both metrics |
| OpenAI GPT-4.1 | 0.793 | 5 | 1.66 | 9 | Yes | Top 8 by COMET-Kiwi |
| Anthropic Claude Sonnet 5 | 0.792 | 6 | 1.66 | 11 | Yes | Top 8 by COMET-Kiwi |
| Google Gemini 3.5 Flash | 0.792 | 7 | 1.64 | 7 | Yes | Top 8 by both metrics |
| Translated Lara 2 | 0.792 | 8 | 1.66 | 10 | Yes | Top 8 by COMET-Kiwi |
| Anthropic Claude Fable 5 | 0.791 | 9 | 1.65 | 8 | Yes | Top 8 by MetricX |
| Google Gemini 3.1 Pro Preview | 0.790 | 10 | 1.61 | 3 | Yes | Top 8 by MetricX |
| Microsoft Translator | 0.782 | 16 | 1.60 | 2 | Yes | Top 8 by MetricX |
| Model | Solution complexity | COMET-Kiwi rank | COMET-Kiwi | MetricX rank | MetricX | Shortlist |
|---|---|---|---|---|---|---|
| OpenAI GPT-5.6 Terra | Adjusted LLM | 1 | 0.797 | 5 | 1.62 | Top 8 by both metrics |
| DeepL Next Gen | Off-the-shelf NMT | 2 | 0.794 | 1 | 1.47 | Top 8 by both metrics |
| Google Gemma 4 | Adjusted LLM | 3 | 0.794 | 4 | 1.61 | Top 8 by both metrics |
| xAI Grok 4.20 (reasoning) | Adjusted LLM | 4 | 0.794 | 6 | 1.64 | Top 8 by both metrics |
| OpenAI GPT-4.1 | Adjusted LLM | 5 | 0.793 | 9 | 1.66 | Top 8 by COMET-Kiwi |
| Anthropic Claude Sonnet 5 | Adjusted LLM | 6 | 0.792 | 11 | 1.66 | Top 8 by COMET-Kiwi |
| Google Gemini 3.5 Flash | Adjusted LLM | 7 | 0.792 | 7 | 1.64 | Top 8 by both metrics |
| Translated Lara 2 | Adjusted NMT | 8 | 0.792 | 10 | 1.66 | Top 8 by COMET-Kiwi |
| Anthropic Claude Fable 5 | Adjusted LLM | 9 | 0.791 | 8 | 1.65 | Top 8 by MetricX |
| Google Gemini 3.1 Pro Preview | Adjusted LLM | 10 | 0.790 | 3 | 1.61 | Top 8 by MetricX |
| OpenAI GPT-5.5 | Adjusted LLM | 11 | 0.789 | 15 | 1.84 | Not shortlisted |
| z.ai GLM-5 | Adjusted LLM | 12 | 0.788 | 14 | 1.69 | Not shortlisted |
| Anthropic Claude Sonnet 4.6 | Adjusted LLM | 13 | 0.787 | 12 | 1.66 | Not shortlisted |
| DeepSeek V3.2 | Adjusted LLM | 14 | 0.787 | 18 | 1.88 | Not shortlisted |
| Anthropic Claude Opus 4.8 | Adjusted LLM | 15 | 0.786 | 13 | 1.67 | Not shortlisted |
| Microsoft Translator | Off-the-shelf NMT | 16 | 0.782 | 2 | 1.60 | Top 8 by MetricX |
| Google Translation LLM | Adjusted NMT | 17 | 0.780 | 23 | 1.98 | Not shortlisted |
| MoonshotAI Kimi K2.5 | Adjusted LLM | 18 | 0.779 | 21 | 1.96 | Not shortlisted |
| Tilde TildeOpen LLM engine | Adjusted NMT | 19 | 0.778 | 17 | 1.87 | Not shortlisted |
| Mistral Large 3 | Adjusted LLM | 20 | 0.776 | 20 | 1.95 | Not shortlisted |
| OpenAI gpt-oss-120b | Adjusted LLM | 21 | 0.776 | 16 | 1.86 | Not shortlisted |
| ModernMT Enterprise Edition | Off-the-shelf NMT | 22 | 0.774 | 19 | 1.90 | Not shortlisted |
| Anthropic Claude Haiku 4.5 | Adjusted LLM | 23 | 0.772 | 24 | 2.03 | Not shortlisted |
| Swiss AI Apertus-8B-2509 | Off-the-shelf LLM | 24 | 0.772 | 26 | 2.15 | Not shortlisted |
| SpeakLeash Bielik-11B-v3.0-Instruct | Adjusted LLM | 25 | 0.768 | 29 | 2.29 | Not shortlisted |
| Cohere Command A Translate | Adjusted LLM | 26 | 0.763 | 22 | 1.97 | Not shortlisted |
| MiniMax M2.5 | Adjusted LLM | 27 | 0.761 | 25 | 2.13 | Not shortlisted |
| Alibaba Qwen3-235B-A22B | Adjusted LLM | 28 | 0.761 | 27 | 2.19 | Not shortlisted |
| Yandex Cloud Translate v2 | Off-the-shelf NMT | 29 | 0.757 | 28 | 2.21 | Not shortlisted |
| Google Cloud Advanced Translation API | Off-the-shelf NMT | 30 | 0.754 | 32 | 2.57 | Not shortlisted |
| Mistral Medium 3.1 | Adjusted LLM | 31 | 0.753 | 30 | 2.31 | Not shortlisted |
| Amazon Translate | Off-the-shelf NMT | 32 | 0.746 | 31 | 2.53 | Not shortlisted |
| Supertext LLM | Off-the-shelf NMT | 33 | 0.746 | 33 | 2.69 | Not shortlisted |
| SYSTRAN PNMT API | Off-the-shelf NMT | 34 | 0.678 | 35 | 3.74 | Not shortlisted |
| AI21 Jamba Large 1.7 | Adjusted LLM | 35 | 0.667 | 34 | 3.42 | Not shortlisted |
| Alibaba Machine Translation API | Off-the-shelf NMT | 36 | 0.645 | 36 | 4.25 | Not shortlisted |
| OpenAI GPT-5.6 Luna | Adjusted LLM | — | 0.793 | — | 1.56 | Same model family as OpenAI GPT-5.6 Terra |
| OpenAI GPT-5.6 Sol | Adjusted LLM | — | 0.793 | — | 1.57 | Same model family as OpenAI GPT-5.6 Terra |
| xAI Grok 4.20 (non-reasoning) | Adjusted LLM | — | 0.790 | — | 1.66 | 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 36 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 automatic metrics and Intento LQA order the shortlist differently. Translations from DeepL Next Gen, first on MetricX and second on COMET-Kiwi, have 36.3 weighted errors per 1,000 words, and those from Google Gemma 4, third on COMET-Kiwi and fourth on MetricX, 51.7. Translations from Microsoft Translator, second on MetricX and 16th on COMET-Kiwi, have the most of the shortlist, 54.4 weighted errors per 1,000 words. The models nearest the cut are OpenAI GPT-5.5, 11th on COMET-Kiwi, and Anthropic Claude Sonnet 4.6, 12th on MetricX.
| Quality level | Cheapest solution that meets every requirement of the level | Cost per million characters |
|---|---|---|
| Basic | OpenAI GPT-5.6 Terra, two-step | $11.46 |
| Standard | OpenAI GPT-5.6 Terra, two-step | $11.46 |
| Advanced | No solution | — |
| Model | US dollars per million characters | Class |
|---|---|---|
| Google Gemma 4 | $0.16 | Open-weight LLMs |
| xAI Grok 4.20 (reasoning) | $1.23 | LLMs |
| OpenAI GPT-4.1 | $3.31 | LLMs |
| Google Gemini 3.5 Flash | $3.94 | LLMs |
| OpenAI GPT-5.6 Terra | $4.84 | LLMs |
| Google Gemini 3.1 Pro Preview | $5.29 | LLMs |
| Anthropic Claude Sonnet 5 | $6.29 | LLMs |
| Microsoft Translator | $10 | MT engines |
| Translated Lara 2 | $24.99 | LLM-based MT engines |
| DeepL Next Gen | $27.5 | LLM-based MT engines |
| Anthropic Claude Fable 5 | $31.75 | LLMs |
| Model | Tokens per character | Class |
|---|---|---|
| Anthropic Claude Opus 4.8 | 0.506 | LLMs |
| Anthropic Claude Fable 5 | 0.505 | LLMs |
| Anthropic Claude Sonnet 5 | 0.505 | LLMs |
| Mistral Large 3 | 0.465 | LLMs |
| Mistral Medium 3.1 | 0.465 | LLMs |
| AI21 Jamba Large 1.7 | 0.452 | LLMs |
| Cohere Command A Translate | 0.406 | LLMs |
| Anthropic Claude Sonnet 4.6 | 0.4 | LLMs |
| Anthropic Claude Haiku 4.5 | 0.398 | LLMs |
| DeepSeek V3.2 | 0.398 | LLMs |
| Alibaba Qwen3-235B-A22B | 0.386 | Open-weight LLMs |
| z.ai GLM-5 | 0.384 | LLMs |
| MoonshotAI Kimi K2.5 | 0.367 | LLMs |
| MiniMax M2.5 | 0.36 | LLMs |
| Google Gemini 3.1 Pro Preview | 0.356 | LLMs |
| Google Gemini 3.5 Flash | 0.355 | LLMs |
| Google Gemma 4 | 0.355 | Open-weight LLMs |
| xAI Grok 4.20 (non-reasoning) | 0.334 | LLMs |
| xAI Grok 4.20 (reasoning) | 0.334 | LLMs |
| OpenAI GPT-5.6 Sol | 0.329 | LLMs |
| OpenAI GPT-5.5 | 0.328 | LLMs |
| OpenAI GPT-5.6 Luna | 0.328 | LLMs |
| OpenAI GPT-5.6 Terra | 0.328 | LLMs |
| OpenAI GPT-4.1 | 0.327 | LLMs |
| Quality level | Fastest solution that meets every requirement of the level | Seconds per 1,000 words |
|---|---|---|
| Basic | OpenAI GPT-5.6 Terra, two-step | 69 |
| Standard | OpenAI GPT-5.6 Terra, two-step | 69 |
| Advanced | No solution | — |
| Model | Seconds per 1,000 words | Class |
|---|---|---|
| Google Cloud Advanced Translation API | 0.1 | MT engines |
| Microsoft Translator | 0.3 | MT engines |
| ModernMT Enterprise Edition | 0.3 | MT engines |
| Google Translation LLM | 0.5 | LLM-based MT engines |
| DeepL Next Gen | 0.7 | LLM-based MT engines |
| SYSTRAN PNMT API | 0.9 | MT engines |
| Amazon Translate | 1.4 | MT engines |
| Translated Lara 2 | 1.5 | LLM-based MT engines |
| Supertext LLM | 1.6 | LLM-based MT engines |
| OpenAI gpt-oss-120b | 2.7 | Open-weight LLMs |
| Yandex Cloud Translate v2 | 3.3 | MT engines |
| Tilde TildeOpen LLM engine | 8.9 | LLM-based MT engines |
| OpenAI GPT-5.6 Luna | 16 | LLMs |
| Alibaba Machine Translation API | 25 | MT engines |
| OpenAI GPT-4.1 | 26 | LLMs |
| OpenAI GPT-5.6 Terra | 31 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 35 | LLMs |
| Swiss AI Apertus-8B-2509 | 45 | Open-weight LLMs |
| Anthropic Claude Haiku 4.5 | 46 | LLMs |
| Mistral Medium 3.1 | 46 | LLMs |
| MiniMax M2.5 | 47 | LLMs |
| Cohere Command A Translate | 64 | LLMs |
| OpenAI GPT-5.6 Sol | 65 | LLMs |
| Mistral Large 3 | 66 | LLMs |
| Google Gemini 3.5 Flash | 67 | LLMs |
| DeepSeek V3.2 | 67 | LLMs |
| AI21 Jamba Large 1.7 | 71 | LLMs |
| SpeakLeash Bielik-11B-v3.0-Instruct | 75 | Open-weight LLMs |
| Anthropic Claude Sonnet 5 | 80 | LLMs |
| Anthropic Claude Fable 5 | 85 | LLMs |
| Anthropic Claude Sonnet 4.6 | 88 | LLMs |
| xAI Grok 4.20 (reasoning) | 95 | LLMs |
| OpenAI GPT-5.5 | 98 | LLMs |
| Google Gemini 3.1 Pro Preview | 103 | LLMs |
| Anthropic Claude Opus 4.8 | 108 | LLMs |
| Alibaba Qwen3-235B-A22B | 126 | Open-weight LLMs |
| MoonshotAI Kimi K2.5 | 142 | LLMs |
| Google Gemma 4 | 195 | Open-weight LLMs |
| z.ai GLM-5 | 235 | 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 Albanian.
Eight of the nine real-time models are MT engines or LLM-based MT engines. None of the three on the shortlist, Microsoft Translator, DeepL Next Gen and Translated Lara 2, meet general translation quality, tag handling or formatting. Of the shortlisted one-step solutions on LLMs, the one on OpenAI GPT-4.1 is the fastest, at 26 seconds per 1,000 words, and the one on Google Gemma 4 the slowest, at 195. The two-step solution (translation + post-edit) on OpenAI GPT-5.6 Terra takes 69 seconds per 1,000 words, against 31 in one step.
| 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 | two-step, with the gender-neutrality agent | 98.2% | ✗ | $62.86 | 158 |
| OpenAI GPT-5.6 Terra | two-step, with the gender-neutrality agent | 98.2% | ✗ | $13.29 | 84 |
For tone of voice in Albanian we check the formal register, with Ju as the form of address. Translations from ten of the 11 shortlisted one-step solutions have 100% of the sentences in it, and those from Microsoft Translator 97.9%. For gender neutrality we check neutral rephrasing, with no default gender, and the other options in the language are a masculine default and paired forms. In one step, translations have a median 5.5 gender-neutrality errors per 1,000 words.
In Albanian the quality-estimation pair of automatic metrics has an agreement of 0.62, above its floor of 0.53, and the similarity pair 0.45, below its floor of 0.46, with no borderline case. The three LQA judges agree on a median of 96% of the sentences across the 11 shortlisted solutions, so the Intento LQA results on this page rest on a review we can rely on.
Leave feedback