Alibaba's Qoder adds fast response mode with 3-second answers and reduced credit consumption
Read more
Pandaily
pandaily.com

Alibaba's Qoder adds fast response mode with 3-second answers and reduced credit consumption

Alibaba's coding AI agent, Qoder, has introduced a new mode called Fast, designed for handling everyday tasks that require prompt results. The company announced on October 10th that this new tier is available in Qoder desktop application version 0.4.4 and later, for both individual and enterprise users, across both Chinese and international versions, according to IT Home.

The Fast mode is characterized by a first response time of approximately three seconds, high output quality, and a credit usage rate of 0.8x. In the Chinese version, Fast is added as an additional option alongside existing ones. Meanwhile, in the international version, it replaces the previous Performance tier rather than being placed next to it.

For international users, this change means the availability of a faster and more economical tier without compromising quality. The first response time has been reduced from about eight seconds to three, and the credit coefficient has decreased from 1.1x to 0.8x, representing an approximate saving of 27%. Qoder asserts that the output quality remains the same as Performance, and users on the old tier are automatically migrated to the new mode without needing any action from them.

Qoder also published first response time metrics for its other tiers. According to the company's own measurements, Fast at three seconds compares to 7.9 seconds for Performance on the international version, 9.8 seconds for Auto, and 13 seconds for Ultimate on the international version. This positions Fast at less than one-third the time of Auto and approximately one-quarter the time of Ultimate.

The company conducted its own comparative test against the fast modes of competing coding tools. In this self-reported testing, which covered ten common developer tasks, Qoder stated that it completed tasks in approximately one-third the time compared to comparable domestic products and approximately one-quarter the time compared to comparable foreign products. The names of the tested products were not disclosed, and the resulting figures have not undergone independent verification.

The example tasks presented in the announcement were relatively simple: drafting a brief travel itinerary took about one second, and converting a daily work log into a plan for the next day took about four seconds. Qoder reported that other products in the Qoder family will soon add the Fast mode, but no exact timeline was provided. This tier is aimed at routine, quick work, while slower tiers like Ultimate remain available for more complex tasks.

Similar stories

Alibaba's XekRung Model Takes First Place in CyberGym Ranking with 88.9% Score
Read more
pandaily.com

Alibaba's XekRung Model Takes First Place in CyberGym Ranking with 88.9% Score

Alibaba's cybersecurity large language model, XekRung, has secured the top spot in the CyberGym ranking, which serves as a benchmark for vulnerability reproduction and is supported by researchers from the University of California, Berkeley. This version, designated as XekRung-1.5-27B-Preview and fine-tuned based on Alibaba's open-source model Qwen3.8-27B, demonstrated a success rate of 88.9% as of September 13th. Chinese tech media reported this result on September 28th, noting it is the first time a model from a Chinese developer has led this ranking.

The CyberGym test includes 1507 real vulnerabilities from 188 open-source projects that were initially discovered by Google's OSS-Fuzz program. In the first level of the task, the model receives a vulnerability description and unpatched code, and must then create an input file in the form of a proof-of-concept that triggers the error in the vulnerable version but not in the patched one.

In the current ranking, the XekRung model surpasses Google's Gemini 3.8 Flash Cyber with a score of 86.3%, OpenAI's GPT-5.5-Cyber with 85.6%, Zhipu AI's GLM-5.3 with 84.5%, and DeepSeek-V4-Pro with 83.3%. Supporters of this benchmark note that evaluations are provided by separate teams, runs are stochastic, and small differences in scores may not reflect significant differences in capabilities.

The model was developed in Alibaba's AGI Security Lab under the guidance of Huang Luntao. According to the team's report, the base model Qwen3.8-27B achieved 54.51% under the same conditions, representing an increase of 34.39 percentage points after additional training. The lab explains this success through supervised fine-tuning on data anonymized from Alibaba's own security operations, reinforcement learning of agents within a general instrumental framework, reward collection based on build, failure, and PoC validation, as well as reusing failed attempts as corrective pairs for training. It is emphasized that no CyberGym tasks, patches, or reference PoCs were used during the training process. The evaluation was conducted on locally deployed FP8 inference with a 256K context window, without pre-installed fuzzing frameworks, and with limited network access according to the list permitted in the benchmark.

Alibaba also emphasizes efficiency, stating that the 27-billion parameter model is 1/27th or 1/370th the size compared to comparable cyber models, which will reduce computational costs for automated vulnerability sorting. The weights of XekRung have not yet been published, and Alibaba has not provided a timeline for external access, although Huang stated that the lab's security technology will be offered more broadly, and future versions will target adversarial intelligence, self-evolution, and agent tasks. A technical report on the 8-billion parameter version built on Qwen was previously released in May.

Popular