The State of Translation Automation 2026 · Languages
What it takes to automate translation from English into Tajik 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 Tajik. |
| Tone of voice | The formal register (Шумо) 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 Tajik, 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. 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.
Three three-step solutions (translation + LQA + repair) meet every Standard requirement in Tajik. Optimizing for quality, cost or latency selects the one on OpenAI GPT-5.6 Sol, with 8.2 total errors per 1,000 words, the lowest cost and the shortest time of the three. Translations from the dearer solution on Anthropic Claude Fable 5 have two major errors and 9.9 weighted errors per 1,000 words, and those from the one on OpenAI GPT-5.6 Sol no major error and 8.2.
None of the solutions under $577 per million characters or under 271 seconds per 1,000 words meet even the Basic requirements in Tajik, and that rules out every real-time solution. Of the three three-step solutions (translation + LQA + repair) that meet every Standard requirement, a cost cap of at most $600 per million characters or a time limit under 316 seconds per 1,000 words leaves only the one on OpenAI GPT-5.6 Sol. A cap from $601 to $638 per million characters adds the one on Google Gemini 3.1 Pro Preview.
The Advanced requirements add gender neutrality and leave two of the three solutions. The three-step solutions (translation + LQA + repair) on OpenAI GPT-5.6 Sol and Anthropic Claude Fable 5 meet them without the gender-neutrality agent, so their cost stays at $577 and $639 per million characters. Their time stays at 271 and 319 seconds per 1,000 words, and their errors do not change. The one on Google Gemini 3.1 Pro Preview falls outside the two best solutions, the ones the Advanced requirements count.
OpenAI GPT-5.6 Sol
three-step
OpenAI GPT-5.6 Sol
three-step
OpenAI GPT-5.6 Sol
three-step
OpenAI GPT-5.6 Sol
three-step
OpenAI GPT-5.6 Sol
three-step
OpenAI GPT-5.6 Sol
three-step
OpenAI GPT-5.6 Sol
three-step
OpenAI GPT-5.6 Sol
three-step
OpenAI GPT-5.6 Sol
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 Sol | 8.2 | $577 | 271 |
| Standard | Tier 2: two or more solutions, none simple | ☆☆☆ | three-step | OpenAI GPT-5.6 Sol | 8.2 | $577 | 271 |
| Advanced | Tier 2: two or more solutions, none simple | ☆☆ | three-step | OpenAI GPT-5.6 Sol | 8.2 | $577 | 271 |
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.
All three three-step solutions (translation + LQA + repair) that meet the Basic requirements in Tajik meet the Standard ones too, since their translations also meet tone of voice, tag handling and formatting. The Advanced requirements leave the two on OpenAI GPT-5.6 Sol and Anthropic Claude Fable 5, without the gender-neutrality agent. At all three levels, optimizing for quality, cost or latency selects OpenAI GPT-5.6 Sol.
| Model | US dollars per million characters | Weighted errors per 1,000 words | Group |
|---|---|---|---|
| OpenAI GPT-4.1 | 3.35 | 71.2 | Misses a Standard requirement |
| Google Gemini 3.5 Flash | 4.4 | 42.2 | Misses a Standard requirement |
| Google Gemini 3.1 Pro Preview | 5.89 | 29.3 | Misses a Standard requirement |
| OpenAI GPT-5.6 Sol | 8.42 | 22.7 | Misses a Standard requirement |
| Anthropic Claude Sonnet 4.6 | 9.83 | 82.7 | Misses a Standard requirement |
| OpenAI GPT-5.5 | 12.33 | 35.6 | Misses a Standard requirement |
| Anthropic Claude Opus 4.8 | 18.09 | 60 | Misses a Standard requirement |
| Google Translation LLM | 21.37 | 86.4 | Misses a Standard requirement |
| Translated Lara 2 | 24.99 | 138.4 | Misses a Standard requirement |
| DeepL Next Gen | 27.5 | 70.5 | Misses a Standard requirement |
| Anthropic Claude Fable 5 | 35.8 | 24.7 | Misses a Standard requirement |
Of the one-step solutions in Tajik with a conclusive Intento LQA result, the one on OpenAI GPT-5.6 Sol has the fewest weighted errors, 22.7 per 1,000 words, for $8.42 per million characters. Below that price, their errors rise as prices fall, up to 71.2 weighted errors per 1,000 words on OpenAI GPT-4.1, for $3.35 per million characters. The three-step solution (translation + LQA + repair) on Google Gemini 3.1 Pro Preview costs more than the one on OpenAI GPT-5.6 Sol, with 10.9 weighted errors per 1,000 words against 8.2.
| Requirement | Median errors per 1,000 words |
|---|---|
| General translation quality | 60 |
| Tone of voice | 0 |
| Tag handling | 0 |
| Gender neutrality | 0 |
The requirements with countable errors. Every solution with its result on each requirement: the explorer.
General translation quality sets the steps in Tajik. The one-step solutions closest to its bar, on OpenAI GPT-5.6 Sol, Anthropic Claude Fable 5 and Google Gemini 3.1 Pro Preview, have 22.7 to 29.3 weighted errors per 1,000 words. Two-step solutions (translation + post-edit) on them have 21.1 to 24.4 weighted errors per 1,000 words. Three-step solutions (translation + LQA + repair) on them have 8.2 to 10.9, and those on OpenAI GPT-4.1 and Anthropic Claude Sonnet 4.6 stay over the bar, at 44.5 and 30.7 weighted errors per 1,000 words.
| Model | Weighted errors per 1,000 words of the shortlisted one-step solutions in Tajik | Class |
|---|---|---|
| Translated Lara 2 | 138.4 | LLM-based MT engines |
| Google Translation LLM | 86.4 | LLM-based MT engines |
| Anthropic Claude Sonnet 4.6 | 82.7 | LLMs |
| OpenAI GPT-4.1 | 71.2 | LLMs |
| DeepL Next Gen | 70.5 | LLM-based MT engines |
| Anthropic Claude Opus 4.8 | 60 | LLMs |
| Google Gemini 3.5 Flash | 42.2 | LLMs |
| OpenAI GPT-5.5 | 35.6 | LLMs |
| Google Gemini 3.1 Pro Preview | 29.3 | LLMs |
| Anthropic Claude Fable 5 | 24.7 | LLMs |
| OpenAI GPT-5.6 Sol | 22.7 | LLMs |
No translation into Tajik: AI21 Jamba Large 1.7 (garbage or near-empty, wrong language, wrong script), Cohere Command A Translate (wrong language, wrong script).
| Rank | Model | MetricX (lower is better) | 83% confidence interval, from | to | Status |
|---|---|---|---|---|---|
| 1 | Google Gemini 3.5 Flash | 1.34 | 1.16 | 1.52 | Shortlisted |
| 2 | Anthropic Claude Fable 5 | 1.42 | 1.25 | 1.59 | Shortlisted |
| 3 | OpenAI GPT-5.6 Sol | 1.44 | 1.25 | 1.62 | Shortlisted |
| 4 | Google Gemini 3.1 Pro Preview | 1.49 | 1.30 | 1.67 | Shortlisted |
| 5 | OpenAI GPT-4.1 | 1.51 | 1.32 | 1.70 | Shortlisted |
| 6 | xAI Grok 4.20 (non-reasoning) | 1.53 | 1.35 | 1.71 | Shortlisted |
| 7 | DeepL Next Gen | 1.53 | 1.34 | 1.72 | Shortlisted |
| 8 | Anthropic Claude Sonnet 4.6 | 1.57 | 1.39 | 1.74 | Shortlisted |
| 9 | OpenAI GPT-5.5 | 1.60 | 1.40 | 1.80 | Shortlisted |
| 10 | Anthropic Claude Opus 4.8 | 1.64 | 1.44 | 1.84 | Shortlisted |
| 11 | z.ai GLM-5 | 1.64 | 1.46 | 1.83 | Not shortlisted |
| 12 | DeepSeek V3.2 | 1.77 | 1.56 | 1.98 | Not shortlisted |
| 13 | Anthropic Claude Sonnet 5 | 1.79 | 1.60 | 1.98 | Not shortlisted |
| 14 | MoonshotAI Kimi K2.5 | 1.88 | 1.68 | 2.07 | Not shortlisted |
| 15 | Google Gemma 4 | 1.89 | 1.68 | 2.11 | Not shortlisted |
| 16 | Google Translation LLM | 1.91 | 1.52 | 2.29 | Shortlisted |
| 17 | Translated Lara 2 | 2.00 | 1.72 | 2.28 | Shortlisted |
| 18 | Yandex Cloud Translate v2 | 2.31 | 2.05 | 2.57 | Not shortlisted |
| 19 | Mistral Large 3 | 2.40 | 2.15 | 2.64 | Not shortlisted |
| 20 | Google Cloud Advanced Translation API | 2.59 | 2.09 | 3.09 | Not shortlisted |
| 21 | MiniMax M2.5 | 2.93 | 2.65 | 3.21 | Not shortlisted |
| 22 | Anthropic Claude Haiku 4.5 | 2.94 | 2.65 | 3.24 | Not shortlisted |
| 23 | OpenAI gpt-oss-120b | 3.44 | 3.00 | 3.89 | Not shortlisted |
| 24 | Alibaba Qwen3-235B-A22B | 3.61 | 3.31 | 3.92 | Not shortlisted |
| 25 | Mistral Medium 3.1 | 3.78 | 3.46 | 4.10 | Not shortlisted |
| 26 | SYSTRAN PNMT API | 4.13 | 3.67 | 4.58 | Not shortlisted |
| 27 | ModernMT Enterprise Edition | 4.59 | 4.07 | 5.10 | Not shortlisted |
| 28 | Swiss AI Apertus-8B-2509 | 7.88 | 6.83 | 8.92 | Not shortlisted |
| Model | Solution complexity | MetricX rank | MetricX | Shortlist |
|---|---|---|---|---|
| Google Gemini 3.5 Flash | Adjusted LLM | 1 | 1.34 | Top 10 by MetricX |
| Anthropic Claude Fable 5 | Adjusted LLM | 2 | 1.42 | Top 10 by MetricX |
| OpenAI GPT-5.6 Sol | Adjusted LLM | 3 | 1.44 | Top 10 by MetricX |
| Google Gemini 3.1 Pro Preview | Adjusted LLM | 4 | 1.49 | Top 10 by MetricX |
| OpenAI GPT-4.1 | Adjusted LLM | 5 | 1.51 | Top 10 by MetricX |
| xAI Grok 4.20 (non-reasoning) | Adjusted LLM | 6 | 1.53 | Top 10 by MetricX |
| DeepL Next Gen | Off-the-shelf NMT | 7 | 1.53 | Top 10 by MetricX |
| Anthropic Claude Sonnet 4.6 | Adjusted LLM | 8 | 1.57 | Top 10 by MetricX |
| OpenAI GPT-5.5 | Adjusted LLM | 9 | 1.60 | Top 10 by MetricX |
| Anthropic Claude Opus 4.8 | Adjusted LLM | 10 | 1.64 | Top 10 by MetricX |
| z.ai GLM-5 | Adjusted LLM | 11 | 1.64 | Not shortlisted |
| DeepSeek V3.2 | Adjusted LLM | 12 | 1.77 | Not shortlisted |
| Anthropic Claude Sonnet 5 | Adjusted LLM | 13 | 1.79 | Not shortlisted |
| MoonshotAI Kimi K2.5 | Adjusted LLM | 14 | 1.88 | Not shortlisted |
| Google Gemma 4 | Adjusted LLM | 15 | 1.89 | Not shortlisted |
| Google Translation LLM | Adjusted NMT | 16 | 1.91 | Real-time place |
| Translated Lara 2 | Adjusted NMT | 17 | 2.00 | Real-time place |
| Yandex Cloud Translate v2 | Off-the-shelf NMT | 18 | 2.31 | Not shortlisted |
| Mistral Large 3 | Adjusted LLM | 19 | 2.40 | Not shortlisted |
| Google Cloud Advanced Translation API | Off-the-shelf NMT | 20 | 2.59 | Not shortlisted |
| MiniMax M2.5 | Adjusted LLM | 21 | 2.93 | Not shortlisted |
| Anthropic Claude Haiku 4.5 | Adjusted LLM | 22 | 2.94 | Not shortlisted |
| OpenAI gpt-oss-120b | Adjusted LLM | 23 | 3.44 | Not shortlisted |
| Alibaba Qwen3-235B-A22B | Adjusted LLM | 24 | 3.61 | Not shortlisted |
| Mistral Medium 3.1 | Adjusted LLM | 25 | 3.78 | Not shortlisted |
| SYSTRAN PNMT API | Off-the-shelf NMT | 26 | 4.13 | Not shortlisted |
| ModernMT Enterprise Edition | Off-the-shelf NMT | 27 | 4.59 | Not shortlisted |
| Swiss AI Apertus-8B-2509 | Off-the-shelf LLM | 28 | 7.88 | Not shortlisted |
| OpenAI GPT-5.6 Luna | Adjusted LLM | — | 1.45 | Same model family as OpenAI GPT-5.6 Sol |
| OpenAI GPT-5.6 Terra | Adjusted LLM | — | 1.54 | Same model family as OpenAI GPT-5.6 Sol |
| xAI Grok 4.20 (reasoning) | Adjusted LLM | — | 1.65 | 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. Ranks among 28 models, with each model family counted once, and the top 10 in bold. Lower is better on MetricX. The scores rank models within this language and are not a requirement result.
MetricX ranks OpenAI GPT-4.1 fifth and OpenAI GPT-5.5 ninth in Tajik. By Intento LQA, the one-step solution on OpenAI GPT-5.5 has 35.6 weighted errors per 1,000 words and the one on OpenAI GPT-4.1 has 71.2. The one on Google Gemini 3.5 Flash, first on MetricX, has 42.2 weighted errors per 1,000 words, against 22.7 for the one on OpenAI GPT-5.6 Sol, third. Nearest the cut in the tail is z.ai GLM-5, 11th on MetricX with a MetricX score of 1.641, against 1.6389 for Anthropic Claude Opus 4.8, the last in the top 10.
| Quality level | Cheapest solution that meets every requirement of the level | Cost per million characters |
|---|---|---|
| Basic | OpenAI GPT-5.6 Sol, three-step | $577 |
| Standard | OpenAI GPT-5.6 Sol, three-step | $577 |
| Advanced | OpenAI GPT-5.6 Sol, three-step | $577 |
| Model | US dollars per million characters | Class |
|---|---|---|
| xAI Grok 4.20 (non-reasoning) | $1.42 | LLMs |
| OpenAI GPT-4.1 | $3.35 | LLMs |
| Google Gemini 3.5 Flash | $4.4 | LLMs |
| Google Gemini 3.1 Pro Preview | $5.89 | LLMs |
| OpenAI GPT-5.6 Sol | $8.42 | LLMs |
| Anthropic Claude Sonnet 4.6 | $9.83 | LLMs |
| OpenAI GPT-5.5 | $12.33 | LLMs |
| Anthropic Claude Opus 4.8 | $18.09 | LLMs |
| Google Translation LLM | $21.37 | 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 | $35.8 | LLMs |
| Model | Tokens per character | Class |
|---|---|---|
| Anthropic Claude Fable 5 | 0.592 | LLMs |
| Anthropic Claude Opus 4.8 | 0.589 | LLMs |
| Anthropic Claude Sonnet 5 | 0.589 | LLMs |
| Anthropic Claude Haiku 4.5 | 0.569 | LLMs |
| Anthropic Claude Sonnet 4.6 | 0.561 | LLMs |
| Mistral Medium 3.1 | 0.541 | LLMs |
| Mistral Large 3 | 0.528 | LLMs |
| Alibaba Qwen3-235B-A22B | 0.504 | Open-weight LLMs |
| DeepSeek V3.2 | 0.474 | LLMs |
| z.ai GLM-5 | 0.458 | LLMs |
| MoonshotAI Kimi K2.5 | 0.455 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 0.414 | LLMs |
| xAI Grok 4.20 (reasoning) | 0.414 | LLMs |
| MiniMax M2.5 | 0.413 | LLMs |
| Google Gemma 4 | 0.399 | Open-weight LLMs |
| Google Gemini 3.1 Pro Preview | 0.397 | LLMs |
| Google Gemini 3.5 Flash | 0.397 | LLMs |
| OpenAI GPT-5.6 Terra | 0.338 | LLMs |
| OpenAI GPT-5.5 | 0.336 | LLMs |
| OpenAI GPT-5.6 Sol | 0.336 | LLMs |
| OpenAI GPT-4.1 | 0.335 | LLMs |
| OpenAI GPT-5.6 Luna | 0.334 | LLMs |
Each step adds to the cost in Tajik. On the three models whose three-step solutions (translation + LQA + repair) meet every Basic requirement, the one-step solutions cost $5.89 to $35.80 per million characters, the two-step ones (translation + post-edit) $19.91 to $80.32 and the three-step ones $577 to $639. The tokenizer of Anthropic Claude Fable 5 produces 0.592 tokens per character of the translation, against 0.336 for OpenAI GPT-5.6 Sol and 0.397 for Google Gemini 3.1 Pro Preview.
| Quality level | Fastest solution that meets every requirement of the level | Seconds per 1,000 words |
|---|---|---|
| Basic | OpenAI GPT-5.6 Sol, three-step | 271 |
| Standard | OpenAI GPT-5.6 Sol, three-step | 271 |
| Advanced | OpenAI GPT-5.6 Sol, three-step | 271 |
| Model | Seconds per 1,000 words | Class |
|---|---|---|
| Google Cloud Advanced Translation API | 0.1 | MT engines |
| ModernMT Enterprise Edition | 0.3 | MT engines |
| DeepL Next Gen | 0.7 | LLM-based MT engines |
| Google Translation LLM | 0.9 | LLM-based MT engines |
| Translated Lara 2 | 2.3 | LLM-based MT engines |
| SYSTRAN PNMT API | 3.8 | MT engines |
| OpenAI gpt-oss-120b | 4.2 | Open-weight LLMs |
| Yandex Cloud Translate v2 | 6.4 | MT engines |
| OpenAI GPT-5.6 Luna | 18 | LLMs |
| xAI Grok 4.20 (non-reasoning) | 22 | LLMs |
| Google Gemini 3.5 Flash | 23 | LLMs |
| OpenAI GPT-4.1 | 48 | LLMs |
| Swiss AI Apertus-8B-2509 | 48 | Open-weight LLMs |
| OpenAI GPT-5.6 Terra | 51 | LLMs |
| xAI Grok 4.20 (reasoning) | 62 | LLMs |
| DeepSeek V3.2 | 63 | LLMs |
| MiniMax M2.5 | 69 | LLMs |
| Anthropic Claude Haiku 4.5 | 69 | LLMs |
| Mistral Medium 3.1 | 70 | LLMs |
| OpenAI GPT-5.5 | 96 | LLMs |
| Mistral Large 3 | 99 | LLMs |
| Google Gemini 3.1 Pro Preview | 104 | LLMs |
| Anthropic Claude Sonnet 5 | 108 | LLMs |
| Anthropic Claude Fable 5 | 118 | LLMs |
| OpenAI GPT-5.6 Sol | 120 | LLMs |
| Anthropic Claude Sonnet 4.6 | 124 | LLMs |
| Anthropic Claude Opus 4.8 | 143 | LLMs |
| Alibaba Qwen3-235B-A22B | 161 | Open-weight LLMs |
| Google Gemma 4 | 177 | Open-weight LLMs |
| z.ai GLM-5 | 288 | LLMs |
| MoonshotAI Kimi K2.5 | 330 | 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 Tajik.
The three real-time models on the Tajik shortlist are the LLM-based MT engines DeepL Next Gen, Google Translation LLM and Translated Lara 2. Their translations meet tone of voice and have 70.5 to 138.4 weighted errors per 1,000 words. On OpenAI GPT-5.6 Sol the time grows with the steps: 120 seconds per 1,000 words in one step, 198 in two (translation + post-edit) and 271 in three (translation + LQA + repair). At every level, that three-step solution is the fastest that meets all the requirements.
| 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 |
|---|---|---|---|---|---|
| OpenAI GPT-5.6 Sol | three-step | 100% | ✓ | $577 | 271 |
| Anthropic Claude Fable 5 | three-step | 100% | ✓ | $639 | 319 |
For Tajik we check the formal register, with Шумо as the form of address, and translations from every shortlisted one-step solution meet tone of voice. For gender we check that a translation does not add a gender the source does not specify, and we allow no default gender. Tajik has no grammatical gender for persons, so a translation is neutral by default, and we check it for gendered terms. In the translations from the two best solutions, 100% of the sentences about people are gender-neutral.
For the 12 shortlisted solutions in Tajik, the median share of sentences on which the three LQA judges agree is 94.1%. With agreement that high, the Intento LQA error counts on this page carry weight. The similarity pair of automatic metrics agrees clearly, not at the border of its threshold. With no COMET-Kiwi scores, the shortlist rests on one metric, MetricX, while the weighted errors on this page come from Intento LQA.
Leave feedback