The State of Translation Automation 2026 · Languages
What it takes to automate translation from English into Telugu 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 Telugu. |
| Tone of voice | The formal register (మీరు) 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 Telugu, 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. 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.
Three solutions meet every Standard requirement in Telugu, and none leads on errors, cost and time at once. Optimizing for quality selects the three-step solution (translation + LQA + repair) on OpenAI GPT-5.5, with 10.9 total errors per 1,000 words, at $591 per million characters and 469 seconds per 1,000 words. Optimizing for cost or latency selects a two-step solution (translation + post-edit), on Google Gemini 3.1 Pro Preview or on Anthropic Claude Fable 5, with 21.5 and 15.5 total errors per 1,000 words.
For the Advanced requirements, the gender-neutrality agent works with the three-step solutions (translation + LQA + repair) on Google Gemini 3.1 Pro Preview and OpenAI GPT-5.5. On OpenAI GPT-5.5 it raises the cost from $591 to $967 per million characters, and total errors fall from 10.9 to 5.6 per 1,000 words. The time changes little on both models: from 280 to 273 seconds per 1,000 words on Google Gemini 3.1 Pro Preview, and from 469 to 466 on OpenAI GPT-5.5.
OpenAI GPT-5.5
three-step
Google Gemini 3.1 Pro Preview
two-step
Anthropic Claude Fable 5
one-step LLM adjusted for Standard requirements
OpenAI GPT-5.5
three-step
Google Gemini 3.1 Pro Preview
two-step
Anthropic Claude Fable 5
two-step
Google Gemini 3.1 Pro Preview
three-step with the gender-neutrality agent
Tied with OpenAI GPT-5.5 (three-step with the gender-neutrality agent).
Google Gemini 3.1 Pro Preview
three-step with the gender-neutrality agent
Google Gemini 3.1 Pro Preview
three-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 | ★ ☆☆ | one-step | Anthropic Claude Fable 5 | 15.2 | $44.47 | 176 |
| Standard | Tier 2: two or more solutions, none simple | ☆☆☆ | two-step | Google Gemini 3.1 Pro Preview | 21.5 | $29.87 | 196 |
| Advanced | Tier 2: two or more solutions, none simple | ☆☆ | with the gender-neutrality agent | Google Gemini 3.1 Pro Preview | 5.6 | $929 | 273 |
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 Standard requirements remove the one-step solution on Anthropic Claude Fable 5 on formatting. Optimizing for latency selects it under the Basic requirements and its two-step version (translation + post-edit) under the Standard ones. Under both, optimizing for quality selects the three-step solution (translation + LQA + repair) on OpenAI GPT-5.5, for cost the two-step one on Google Gemini 3.1 Pro Preview. Under the Advanced ones, all three select the three-step one on Google Gemini 3.1 Pro Preview with the gender-neutrality agent, tied for quality with the one on OpenAI GPT-5.5.
| Model | US dollars per million characters | Weighted errors per 1,000 words | Group |
|---|---|---|---|
| Google Gemma 4 | 0.15 | 67.6 | Misses a Standard requirement |
| OpenAI GPT-5.6 Luna | 0.51 | 28 | Misses a Standard requirement |
| Mistral Large 3 | 2.2 | 72.2 | Misses a Standard requirement |
| Google Gemini 3.5 Flash | 3.82 | 45.8 | Misses a Standard requirement |
| Google Gemini 3.1 Pro Preview | 5.04 | 29.7 | Misses a Standard requirement |
| Anthropic Claude Sonnet 5 | 8.89 | 40.5 | Misses a Standard requirement |
| Microsoft Translator | 10 | 157.9 | Misses a Standard requirement |
| OpenAI GPT-5.5 | 12.92 | 29 | Misses a Standard requirement |
| Anthropic Claude Opus 4.8 | 21.95 | 20.8 | Misses a Standard requirement |
| Translated Lara 2 | 24.99 | 49.1 | Misses a Standard requirement |
| DeepL Next Gen | 27.5 | 41.9 | Misses a Standard requirement |
| Anthropic Claude Fable 5 | 44.47 | 14.5 | Misses a Standard requirement |
In one step in Telugu, the dearest solution, on Anthropic Claude Fable 5, at $44.47 per million characters, also has the fewest weighted errors, 14.5 per 1,000 words. Yet the one on OpenAI GPT-5.6 Luna, at $0.51 per million characters, has 28.0 weighted errors per 1,000 words, fewer than those on OpenAI GPT-5.5 and Google Gemini 3.1 Pro Preview at $12.92 and $5.04. The cost and errors front of the solutions that meet every Standard requirement holds three solutions, all on LLMs.
| Requirement | Median errors per 1,000 words |
|---|---|
| General translation quality | 41.2 |
| Gender neutrality | 1.9 |
| Tag handling | 1.3 |
| Tone of voice | 0 |
The requirements with countable errors. Every solution with its result on each requirement: the explorer.
In Telugu, the shortlisted one-step solutions have a median of 41.2 weighted errors per 1,000 words, over the bar. Translations from four out of the 12 have 99.2% of the tags right, against 100% for tag handling. The two-step solution (translation + post-edit) on Google Gemini 3.1 Pro Preview has 19.8 weighted errors per 1,000 words, under the bar, against 29.7 in one step. The three-step solution (translation + LQA + repair) on OpenAI GPT-5.5 has 9.2 weighted errors per 1,000 words, against 29.0 in one step and 27.7 in two.
| Model | Weighted errors per 1,000 words of the shortlisted one-step solutions in Telugu | Class |
|---|---|---|
| Microsoft Translator | 157.9 | MT engines |
| Mistral Large 3 | 72.2 | LLMs |
| Google Gemma 4 | 67.6 | Open-weight LLMs |
| Translated Lara 2 | 49.1 | LLM-based MT engines |
| Google Gemini 3.5 Flash | 45.8 | LLMs |
| DeepL Next Gen | 41.9 | LLM-based MT engines |
| Anthropic Claude Sonnet 5 | 40.5 | LLMs |
| Google Gemini 3.1 Pro Preview | 29.7 | LLMs |
| OpenAI GPT-5.5 | 29 | LLMs |
| OpenAI GPT-5.6 Luna | 28 | LLMs |
| Anthropic Claude Opus 4.8 | 20.8 | LLMs |
| Anthropic Claude Fable 5 | 14.5 | LLMs |
No translation into Telugu: AI21 Jamba Large 1.7 (cut off part-way, garbage or near-empty), MiniMax M2.5 (cut off part-way, wrong language, wrong script, garbage or near-empty).
The one-step solutions on the seven shortlisted LLMs that meet our language-support criteria in Telugu have 14.5 to 45.8 weighted errors per 1,000 words. Those on the LLM-based MT engines DeepL Next Gen and Translated Lara 2 have 41.9 and 49.1, and the one on the MT engine Microsoft Translator 157.9 weighted errors per 1,000 words, with 61 major errors. The two LLMs that miss the criteria are Google Gemma 4, with 5 critical errors and 67.6 weighted errors per 1,000 words, and Mistral Large 3, with 72.2.
| Model | COMET-Kiwi | Rank on COMET-Kiwi | MetricX | Rank on MetricX | Shortlisted | Shortlist place |
|---|---|---|---|---|---|---|
| Anthropic Claude Sonnet 4.6 | 0.613 | 9 | 1.41 | 11 | No | Not shortlisted |
| z.ai GLM-5 | 0.613 | 11 | 1.49 | 14 | No | Not shortlisted |
| DeepSeek V3.2 | 0.612 | 13 | 1.50 | 15 | No | Not shortlisted |
| xAI Grok 4.20 (reasoning) | 0.612 | 15 | 1.62 | 16 | No | Not shortlisted |
| OpenAI GPT-4.1 | 0.610 | 16 | 1.47 | 12 | No | Not shortlisted |
| MoonshotAI Kimi K2.5 | 0.607 | 18 | 1.83 | 21 | No | Not shortlisted |
| OpenAI gpt-oss-120b | 0.606 | 19 | 1.70 | 18 | No | Not shortlisted |
| Mistral Medium 3.1 | 0.606 | 20 | 1.74 | 19 | No | Not shortlisted |
| ModernMT Enterprise Edition | 0.604 | 21 | 1.75 | 20 | No | Not shortlisted |
| Google Translation LLM | 0.603 | 22 | 1.90 | 22 | No | Not shortlisted |
| Anthropic Claude Haiku 4.5 | 0.601 | 23 | 2.16 | 23 | No | Not shortlisted |
| Yandex Cloud Translate v2 | 0.598 | 24 | 2.57 | 27 | No | Not shortlisted |
| Amazon Translate | 0.596 | 25 | 2.39 | 26 | No | Not shortlisted |
| Alibaba Qwen3-235B-A22B | 0.595 | 26 | 2.17 | 24 | No | Not shortlisted |
| Google Cloud Advanced Translation API | 0.591 | 27 | 2.18 | 25 | No | Not shortlisted |
| Swiss AI Apertus-8B-2509 | 0.584 | 28 | 2.69 | 28 | No | Not shortlisted |
| Cohere Tiny Aya | 0.548 | 29 | 3.73 | 29 | No | Not shortlisted |
| AppTek NMT | 0.518 | 30 | 5.42 | 30 | No | Not shortlisted |
| Alibaba Machine Translation API | 0.472 | 31 | 8.72 | 31 | No | Not shortlisted |
| Cohere Command A Translate | 0.369 | 32 | 11.05 | 32 | No | Not shortlisted |
| Translated Lara 2 | 0.616 | 1 | 1.48 | 13 | Yes | Top 8 by COMET-Kiwi |
| Google Gemini 3.5 Flash | 0.616 | 2 | 1.29 | 4 | Yes | Top 8 by both metrics |
| Anthropic Claude Fable 5 | 0.615 | 3 | 1.26 | 3 | Yes | Top 8 by both metrics |
| DeepL Next Gen | 0.615 | 4 | 1.32 | 6 | Yes | Top 8 by both metrics |
| Google Gemma 4 | 0.615 | 5 | 1.21 | 1 | Yes | Top 8 by both metrics |
| Anthropic Claude Opus 4.8 | 0.615 | 6 | 1.38 | 10 | Yes | Top 8 by COMET-Kiwi |
| Google Gemini 3.1 Pro Preview | 0.615 | 7 | 1.22 | 2 | Yes | Top 8 by both metrics |
| Anthropic Claude Sonnet 5 | 0.614 | 8 | 1.37 | 9 | Yes | Top 8 by COMET-Kiwi |
| OpenAI GPT-5.5 | 0.613 | 10 | 1.31 | 5 | Yes | Top 8 by MetricX |
| OpenAI GPT-5.6 Luna | 0.613 | 12 | 1.37 | 8 | Yes | Top 8 by MetricX |
| Mistral Large 3 | 0.612 | 14 | 1.36 | 7 | Yes | Top 8 by MetricX |
| Microsoft Translator | 0.608 | 17 | 1.67 | 17 | Yes | Real-time place |
| Model | Solution complexity | COMET-Kiwi rank | COMET-Kiwi | MetricX rank | MetricX | Shortlist |
|---|---|---|---|---|---|---|
| Translated Lara 2 | Adjusted NMT | 1 | 0.616 | 13 | 1.48 | Top 8 by COMET-Kiwi |
| Google Gemini 3.5 Flash | Adjusted LLM | 2 | 0.616 | 4 | 1.29 | Top 8 by both metrics |
| Anthropic Claude Fable 5 | Adjusted LLM | 3 | 0.615 | 3 | 1.26 | Top 8 by both metrics |
| DeepL Next Gen | Off-the-shelf NMT | 4 | 0.615 | 6 | 1.32 | Top 8 by both metrics |
| Google Gemma 4 | Adjusted LLM | 5 | 0.615 | 1 | 1.21 | Top 8 by both metrics |
| Anthropic Claude Opus 4.8 | Adjusted LLM | 6 | 0.615 | 10 | 1.38 | Top 8 by COMET-Kiwi |
| Google Gemini 3.1 Pro Preview | Adjusted LLM | 7 | 0.615 | 2 | 1.22 | Top 8 by both metrics |
| Anthropic Claude Sonnet 5 | Adjusted LLM | 8 | 0.614 | 9 | 1.37 | Top 8 by COMET-Kiwi |
| Anthropic Claude Sonnet 4.6 | Adjusted LLM | 9 | 0.613 | 11 | 1.41 | Not shortlisted |
| OpenAI GPT-5.5 | Adjusted LLM | 10 | 0.613 | 5 | 1.31 | Top 8 by MetricX |
| z.ai GLM-5 | Adjusted LLM | 11 | 0.613 | 14 | 1.49 | Not shortlisted |
| OpenAI GPT-5.6 Luna | Adjusted LLM | 12 | 0.613 | 8 | 1.37 | Top 8 by MetricX |
| DeepSeek V3.2 | Adjusted LLM | 13 | 0.612 | 15 | 1.50 | Not shortlisted |
| Mistral Large 3 | Adjusted LLM | 14 | 0.612 | 7 | 1.36 | Top 8 by MetricX |
| xAI Grok 4.20 (reasoning) | Adjusted LLM | 15 | 0.612 | 16 | 1.62 | Not shortlisted |
| OpenAI GPT-4.1 | Adjusted LLM | 16 | 0.610 | 12 | 1.47 | Not shortlisted |
| Microsoft Translator | Off-the-shelf NMT | 17 | 0.608 | 17 | 1.67 | Real-time place |
| MoonshotAI Kimi K2.5 | Adjusted LLM | 18 | 0.607 | 21 | 1.83 | Not shortlisted |
| OpenAI gpt-oss-120b | Adjusted LLM | 19 | 0.606 | 18 | 1.70 | Not shortlisted |
| Mistral Medium 3.1 | Adjusted LLM | 20 | 0.606 | 19 | 1.74 | Not shortlisted |
| ModernMT Enterprise Edition | Off-the-shelf NMT | 21 | 0.604 | 20 | 1.75 | Not shortlisted |
| Google Translation LLM | Adjusted NMT | 22 | 0.603 | 22 | 1.90 | Not shortlisted |
| Anthropic Claude Haiku 4.5 | Adjusted LLM | 23 | 0.601 | 23 | 2.16 | Not shortlisted |
| Yandex Cloud Translate v2 | Off-the-shelf NMT | 24 | 0.598 | 27 | 2.57 | Not shortlisted |
| Amazon Translate | Off-the-shelf NMT | 25 | 0.596 | 26 | 2.39 | Not shortlisted |
| Alibaba Qwen3-235B-A22B | Adjusted LLM | 26 | 0.595 | 24 | 2.17 | Not shortlisted |
| Google Cloud Advanced Translation API | Off-the-shelf NMT | 27 | 0.591 | 25 | 2.18 | Not shortlisted |
| Swiss AI Apertus-8B-2509 | Off-the-shelf LLM | 28 | 0.584 | 28 | 2.69 | Not shortlisted |
| Cohere Tiny Aya | Off-the-shelf LLM | 29 | 0.548 | 29 | 3.73 | Not shortlisted |
| AppTek NMT | Off-the-shelf NMT | 30 | 0.518 | 30 | 5.42 | Not shortlisted |
| Alibaba Machine Translation API | Off-the-shelf NMT | 31 | 0.472 | 31 | 8.72 | Not shortlisted |
| Cohere Command A Translate | Adjusted LLM | 32 | 0.369 | 32 | 11.05 | Not shortlisted |
| OpenAI GPT-5.6 Sol | Adjusted LLM | — | 0.611 | — | 1.30 | Same model family as OpenAI GPT-5.6 Luna |
| OpenAI GPT-5.6 Terra | Adjusted LLM | — | 0.610 | — | 1.45 | Same model family as OpenAI GPT-5.6 Luna |
| xAI Grok 4.20 (non-reasoning) | Adjusted LLM | — | 0.603 | — | 1.62 | 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 32 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.
By Intento LQA, the one-step solutions on Translated Lara 2 and Google Gemma 4, first on one metric each, have 49.1 and 67.6 weighted errors per 1,000 words in Telugu. The one on Anthropic Claude Fable 5, third on both metrics, has the fewest, 14.5 weighted errors per 1,000 words. MetricX alone put OpenAI GPT-5.5 on the list, at rank 5, and the three-step solution (translation + LQA + repair) on it meets every Standard requirement. Nearest the cut in the tail is Anthropic Claude Sonnet 4.6, 9th on COMET-Kiwi and 11th on MetricX.
| Quality level | Cheapest solution that meets every requirement of the level | Cost per million characters |
|---|---|---|
| Basic | Google Gemini 3.1 Pro Preview, two-step | $29.87 |
| Standard | Google Gemini 3.1 Pro Preview, two-step | $29.87 |
| Advanced | Google Gemini 3.1 Pro Preview, three-step with the gender-neutrality agent | $929 |
| Model | US dollars per million characters | Class |
|---|---|---|
| Google Gemma 4 | $0.15 | Open-weight LLMs |
| OpenAI GPT-5.6 Luna | $0.51 | LLMs |
| Mistral Large 3 | $2.2 | LLMs |
| Google Gemini 3.5 Flash | $3.82 | LLMs |
| Google Gemini 3.1 Pro Preview | $5.04 | LLMs |
| Anthropic Claude Sonnet 5 | $8.89 | LLMs |
| Microsoft Translator | $10 | MT engines |
| OpenAI GPT-5.5 | $12.92 | LLMs |
| Anthropic Claude Opus 4.8 | $21.95 | LLMs |
| Translated Lara 2 | $24.99 | LLM-based MT engines |
| DeepL Next Gen | $27.5 | LLM-based MT engines |
| Anthropic Claude Fable 5 | $44.47 | LLMs |
| Model | Tokens per character | Class |
|---|---|---|
| z.ai GLM-5 | 1.392 | LLMs |
| Cohere Command A Translate | 1.372 | LLMs |
| Mistral Large 3 | 1.352 | LLMs |
| Mistral Medium 3.1 | 1.344 | LLMs |
| Alibaba Qwen3-235B-A22B | 1.203 | Open-weight LLMs |
| Anthropic Claude Sonnet 5 | 0.789 | LLMs |
| Anthropic Claude Fable 5 | 0.782 | LLMs |
| Anthropic Claude Opus 4.8 | 0.776 | LLMs |
| Anthropic Claude Haiku 4.5 | 0.775 | LLMs |
| Anthropic Claude Sonnet 4.6 | 0.762 | LLMs |
| xAI Grok 4.20 (reasoning) | 0.731 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 0.728 | LLMs |
| MoonshotAI Kimi K2.5 | 0.701 | LLMs |
| DeepSeek V3.2 | 0.661 | LLMs |
| OpenAI GPT-5.6 Terra | 0.377 | LLMs |
| OpenAI GPT-5.6 Sol | 0.375 | LLMs |
| OpenAI GPT-5.5 | 0.371 | LLMs |
| OpenAI GPT-4.1 | 0.37 | LLMs |
| OpenAI GPT-5.6 Luna | 0.37 | LLMs |
| Cohere Tiny Aya | 0.362 | LLMs |
| Google Gemini 3.1 Pro Preview | 0.354 | LLMs |
| Google Gemma 4 | 0.352 | Open-weight LLMs |
| Google Gemini 3.5 Flash | 0.35 | LLMs |
Tokenizers of the shortlisted Google and OpenAI models produce 0.35 to 0.371 tokens per character in Telugu, those of the three shortlisted Anthropic Claude models 0.776 to 0.789. Even so, the one-step solution on Anthropic Claude Sonnet 5 costs $8.89 per million characters, less than the $12.92 of the one on OpenAI GPT-5.5. On Google Gemini 3.1 Pro Preview, cost follows the steps: $5.04 per million characters in one step, $29.87 in two steps (translation + post-edit) and $555 in three (translation + LQA + repair).
| Quality level | Fastest solution that meets every requirement of the level | Seconds per 1,000 words |
|---|---|---|
| Basic | Anthropic Claude Fable 5, one-step LLM adjusted for Standard requirements | 176 |
| Standard | Anthropic Claude Fable 5, two-step | 176 |
| Advanced | Google Gemini 3.1 Pro Preview, three-step with the gender-neutrality agent | 273 |
| 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 |
| Google Translation LLM | 0.6 | LLM-based MT engines |
| DeepL Next Gen | 0.8 | LLM-based MT engines |
| Amazon Translate | 1.1 | MT engines |
| Translated Lara 2 | 2.4 | LLM-based MT engines |
| OpenAI gpt-oss-120b | 2.7 | Open-weight LLMs |
| Yandex Cloud Translate v2 | 3.5 | MT engines |
| Google Gemini 3.5 Flash | 17 | LLMs |
| OpenAI GPT-5.6 Luna | 17 | LLMs |
| Cohere Tiny Aya | 19 | LLMs |
| Alibaba Machine Translation API | 27 | MT engines |
| OpenAI GPT-4.1 | 35 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 38 | LLMs |
| AppTek NMT | 40 | MT engines |
| Swiss AI Apertus-8B-2509 | 52 | Open-weight LLMs |
| Mistral Medium 3.1 | 64 | LLMs |
| Anthropic Claude Haiku 4.5 | 73 | LLMs |
| OpenAI GPT-5.6 Terra | 76 | LLMs |
| DeepSeek V3.2 | 77 | LLMs |
| Cohere Command A Translate | 82 | LLMs |
| Mistral Large 3 | 87 | LLMs |
| xAI Grok 4.20 (reasoning) | 95 | LLMs |
| Google Gemini 3.1 Pro Preview | 96 | LLMs |
| Anthropic Claude Sonnet 4.6 | 138 | LLMs |
| Anthropic Claude Sonnet 5 | 146 | LLMs |
| OpenAI GPT-5.5 | 165 | LLMs |
| Anthropic Claude Opus 4.8 | 174 | LLMs |
| Anthropic Claude Fable 5 | 176 | LLMs |
| OpenAI GPT-5.6 Sol | 178 | LLMs |
| Google Gemma 4 | 185 | Open-weight LLMs |
| Alibaba Qwen3-235B-A22B | 377 | Open-weight LLMs |
| MoonshotAI Kimi K2.5 | 388 | LLMs |
| z.ai GLM-5 | 524 | 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 Telugu.
| 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 |
|---|---|---|---|---|---|
| Google Gemini 3.1 Pro Preview | three-step, with the gender-neutrality agent | 100% | ✓ | $929 | 273 |
| OpenAI GPT-5.5 | three-step, with the gender-neutrality agent | 100% | ✓ | $967 | 466 |
For Telugu we check the formal register, with మీరు, and gender-neutral wording through rephrasing, the honorific plural and neutral nouns, with no default gender. The other options in Telugu include the gendered third-person singular. With the gender-neutrality agent, 100% of the sentences about people in the translations from the three-step solutions (translation + LQA + repair) on Google Gemini 3.1 Pro Preview and OpenAI GPT-5.5 are gender-neutral, and both have 5.6 total errors per 1,000 words.
For the 12 shortlisted solutions in Telugu, the median share of sentences on which the three LQA judges agree is 94.7%. With agreement that high, the Intento LQA error counts on this page carry weight.
Leave feedback