How Alibaba’s Tongyi Lingma AI Programmer Supercharges Java Development with QwQ‑32B

Alibaba Cloud has added a model‑selection dropdown to Tongyi Lingma, introducing DeepSeek V3, DeepSeek R1, Qwen2.5‑Max and the newly released QwQ‑32B, expanding the platform’s "luxury lineup" of inference models.

QwQ‑32B, open‑sourced by Alibaba Cloud, achieves a qualitative leap in mathematics, code, and general abilities through large‑scale reinforcement learning, matching DeepSeek‑R1 performance while reducing deployment costs enough for consumer‑grade GPUs.

Developers can switch models directly in the intelligent Q&A window to enjoy personalized services.

QwQ‑32B is freely downloadable for commercial use and consistently outperforms OpenAI‑o1‑mini in benchmark tests such as AIME24 (math), LiveCodeBench (code), LiveBench, IFEval, and BFCL, often surpassing DeepSeek‑R1.

The AI programmer, launched in January, supports VS Code and JetBrains IDEs, enabling end‑to‑end dialogue‑driven development from requirement to code, including multi‑file modifications.

To demonstrate, a CMS notice‑management feature was built:

Designing a MySQL table sys_cms_notice via natural‑language prompt.

Generating controller, service, mapper, domain, and MyBatis XML code automatically.

Creating front‑end add, edit, and list HTML pages by referencing existing templates.

Installation of the Tongyi Lingma plugin in IntelliJ IDEA is straightforward via the Plugins market.

Cross‑language programming was showcased by adding a prompt input and a "Generate" button to the add‑page, which calls Alibaba Cloud’s Bailei DeepSeek‑R1 model to produce announcement content.

// Front‑end generate method
function genera() {
    var multiLineText = $("#multiLineText").val();
    if (!multiLineText) { $.modal.alertWarning("请输入关键词prompt。"); return; }
    $.ajax({
        type: "POST",
        url: ctx + "system/notice/generateContent",
        data: { multiLineText: multiLineText },
        dataType: 'json',
        success: function(result) {
            if (result.code == web_status.SUCCESS) {
                console.log(result.data);
                $('.summernote').summernote('code', result.data);
            } else { $.modal.alertError(result.msg); }
        },
        error: function() { $.modal.alertWarning("生成内容失败。"); }
    });
}

@PostMapping("/generateContent")
@ResponseBody
public AjaxResult generateContent(String multiLineText) {
    String apiKey = "sk-..."; // Replace with actual key
    String url = "https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions";
    JSONObject requestBody = new JSONObject();
    requestBody.put("model", "deepseek-r1");
    JSONArray messages = new JSONArray();
    JSONObject message = new JSONObject();
    message.put("role", "user");
    message.put("content", multiLineText);
    messages.add(message);
    requestBody.put("messages", messages);
    try {
        HttpURLConnection conn = (HttpURLConnection) new URL(url).openConnection();
        conn.setRequestMethod("POST");
        conn.setRequestProperty("Authorization", apiKey);
        conn.setRequestProperty("Content-Type", "application/json");
        conn.setDoOutput(true);
        try (OutputStream os = conn.getOutputStream()) {
            os.write(requestBody.toString().getBytes(StandardCharsets.UTF_8));
        }
        int code = conn.getResponseCode();
        if (code == HttpURLConnection.HTTP_OK) {
            BufferedReader br = new BufferedReader(new InputStreamReader(conn.getInputStream(), StandardCharsets.UTF_8));
            StringBuilder resp = new StringBuilder();
            String line;
            while ((line = br.readLine()) != null) { resp.append(line.trim()); }
            JSONObject parse = (JSONObject) JSON.parse(resp.toString());
            JSONArray choices = parse.getJSONArray("choices");
            StringBuilder sb = new StringBuilder();
            for (Object c : choices) {
                JSONObject obj = (JSONObject) c;
                sb.append(obj.getJSONObject("message").getString("reasoning_content"));
            }
            return AjaxResult.success("success", sb.toString());
        } else {
            return AjaxResult.error("Error: " + code);
        }
    } catch (Exception e) {
        e.printStackTrace();
        return AjaxResult.error("Exception: " + e.getMessage());
    }
}

Developers reported that while the generated code is largely functional (≈95% accuracy), integration often requires manual adjustments such as adding missing page‑navigation methods, aligning with project style, and handling UI quirks like assigning content to rich‑text editors.

Compared with Tongyi Lingma 1.0, version 2.0 supports multiple models, faster generation, richer context types (#file, #codeChanges, #gitCommit, #teamDocs, #image), and a more capable AI programmer that can handle multi‑file changes and tool usage.

Overall, Tongyi Lingma 2.0 demonstrates strong cross‑language programming abilities, high‑quality code generation, and effective error‑resolution assistance, though developers should still provide clear error details and may need to reference existing code to maintain consistency.