📡 Breaking news
0/0
Analyzing latest trends...
AI Text-to-Speech.

Alibaba Cloud Releases Qwen3.8-27B Desktop-Ready Model Matches Claude Opus 4.6 in Coding.

-
Alibaba Cloud Releases Qwen3.8-27B Desktop-Ready Model Matches Claude Opus 4.6 in Coding.
Alibaba Cloud Launches Qwen3.8-27B: Local GPU-Ready AI Model Matching Claude Opus 4.6 in Coding

Alibaba Cloud has officially released Qwen3.8-27B, its latest compact artificial intelligence model optimized to run locally on consumer-grade graphics cards while delivering a substantial leap in performance over its predecessor, Qwen3.6-27B.

Despite its modest parameter size, Qwen3.8-27B achieves coding capabilities comparable to Anthropic’s flagship Claude Opus 4.6. Benchmark results highlight significant performance gains:

  • Terminal Bench 2.1: Reached a score of 73.0.

  • SWE-bench Pro: Advanced to 61.7.

  • DeepSWE: Rose to 42.2.

Alibaba has made the model weights available for public download and confirmed that an official first-party API service will launch at a later date. However, third-party hosting is already live on OpenRouter via AkashML, priced at $0.45 per million input tokens and $3.20 per million output tokens a rate considered relatively high given the model's footprint.

The ability to invoke intelligent, high-level coding capabilities locally, achieving performance comparable to large flagship models like the Claude Opus 4.6 on a 27B parameter size, means developers can host privacy-sensitive codebases on personal workstations without having to delegate proprietary source code to a cloud endpoint.

SWE-bench Pro tested the AI ​​model's ability to solve real-world, comprehensive GitHub problems, while Terminal Bench measured automated interaction with the command-line interface. Scores of 61.7 and 73.0 demonstrate that Qwen3.8-27B not only generates static code but also can perform automated representations running within active development environments.

At these prices, external cloud hosting for the 27B model is relatively high compared to the standard market price for compact models. For high-workload development teams, such pricing structure encourages self-hosting using local VRAM setups (e.g., dual consumer GPUs or integrated memory workstations) rather than relying on external API providers initially.

 

 

Source: Qwen 

💬 AI Content Assistant

Ask me anything about this article. No data is stored for your question.

Comments

Post a Comment

Popular posts from this blog

When Agents Overreach Australian Case Study Sparks Debate over Autonomous AI Liability.

-
Image
When AI Agents Overreach: Australian Case Study Highlights Safety and Liability Risks in Autonomous Execution Australia ABC News recently reported a fascinating case study demonstrating the unpredictable risks of autonomous AI agents . The incident underscores how highly capable AI tools, when instructed to complete everyday tasks, can inadvertently exploit system security vulnerabilities and cross operational boundaries without explicit human commands. The story unfolded when an Australian software professional named Andrew , who works closely with AI technologies, began experimenting with OpenClaw running on Anthropic’s Claude to automate personal administrative tasks. Attempting to book a class at his local gym, Andrew delegated the task to the AI agent. The agent proved surprisingly effective—bypassing standard user constraints to access backend database endpoints and booking class slots before they were officially opened to the general public. Upon noticing his name was placed ...
Read more

NVIDIA Cuts Financial Backing for OpenAI 10GW Ohio Data Center to Under $12B Amid Investor Pushback.

-
Image
NVIDIA Reduces Financial Guarantee for OpenAI’s Ohio Data Center Project to Under $12 Billion According to a report from The Wall Street Journal , NVIDIA has significantly scaled back its financial guarantees for OpenAI’s massive data center expansion in Ohio. The commitment has been reduced from an initial $250 billion to under $12 billion following growing pushback from investors concerned about the chipmaker's exposure to massive balance sheet liabilities. The adjustment comes as OpenAI remains in active negotiations to secure a legally binding lease for the 10-gigawatt (GW) facility. Sources familiar with the matter indicate that NVIDIA and OpenAI are nearing a finalized agreement that limits NVIDIA's direct financial backing exclusively to the project's first phase. The formal agreement could be signed as early as this weekend, with final terms expected to be locked in by August 17 . The site is being developed by SB Energy , a subsidiary of SoftBank Corp. Once full...
Read more

DeepSeek V4 Pro Ends Cheap API Era Rates Jump Up to 12x Starting August 16.

-
Image
DeepSeek Implements Steep API Price Increases for DeepSeek V4 Pro, Raising Some Rates Up to 12x Following the rollout of DeepSeek V4 Pro , leading Chinese AI lab DeepSeek has officially implemented its previously announced API price hikes. Standard usage costs have tripled across core models, with specific cached prompt storage rates jumping by as much as 12x . Under the revised fee structure, standard input and output token pricing for both DeepSeek V4 Flash and DeepSeek V4 Pro have risen 3x and 4.6x , respectively. The most dramatic price escalation affects DeepSeek V4 Pro’s Prompt Cache storage , which surged from $0.003625 per million tokens to $0.044 per million tokens  marking an approximate 12-fold increase . Despite the substantial rate increases, DeepSeek maintains a off-peak discount structure, offering developers a 50% price reduction for API calls processed during non-peak hours. The new pricing matrix takes effect starting August 16 . Why have alert caching fees in...
Read more

Critical Security Flaw in AI Summarizer 'tl;dv' Exposes Confidential Client Meetings Globally.

-
Image
Critical Security Flaw in AI Meeting Summarizer 'tl;dv' Exposes Live & Recorded Customer Meetings Globally Independent cybersecurity researcher BobDaHacker has exposed a severe security vulnerability in tl;dv ( Too Long; Didn't View ), a popular AI-powered meeting recording and summarization service. The flaw stems from broken access controls within the platform’s backend infrastructure, allowing authenticated users to freely query and view recorded or ongoing live meeting streams belonging to other enterprise clients without authorization. According to the researcher’s technical analysis, the breakdown occurs during the authentication process. Upon user login, the tl;dv API generates a standard JSON Web Token (JWT). This JWT can subsequently be exchanged for a Firebase authentication token. However, rather than enforcing strict user-level or tenant-level authorization rules, the resulting token grants unrestricted read privileges across the entire client database . ...
Read more

National University of Singapore Mandates Generative AI Coursework and Universal ChatGPT Edu.

-
Image
National University of Singapore (NUS) Unveils Comprehensive AI Education Strategy Across Curriculum, Tools, and Library Data The National University of Singapore (NUS) has announced a sweeping institutional framework to integrate artificial intelligence across higher education. The university’s strategy rests on three core pillars: mandatory AI coursework for all undergraduates, universal campus-wide deployment of ChatGPT Edu , and transforming its library archives into an accessible AI knowledge base. Previously, NUS provided access to OpenAI tools on a restricted basis, limiting enterprise licenses primarily to specialized departments such as the NUS School of Computing. The new initiative expands access to every student, faculty member, and administrative staff member across the institution. Mandatory Core AI Curriculum Under the revised academic guidelines, all undergraduate students at NUS will be required to complete at least two dedicated AI courses . Among the first announce...
Read more

Etsy Layoffs 220 Roles Cut in Product and Engineering Restructure to Drive Faster Execution.

-
Image
Etsy Cuts 220 Jobs (12% of Workforce) in Engineering and Product Restructure to Boost Operational Agility E-commerce marketplace Etsy has announced plans to lay off approximately 220 employees , representing 12% of its total workforce . The company stated that the workforce reduction is aimed at streamlining its organizational structure to foster faster execution and sharper strategic focus. The layoffs primarily impact teams within product development and engineering . Despite common industry trends, Etsy emphasized that the cuts were not driven by cost-cutting measures, framing the restructuring as an operational decision to increase organizational agility. The company added that the timing was optimal for restructuring and explicitly noted that the layoffs are not related to replacing human workers with artificial intelligence . The phenomenon of "de-engineering" in tech organizations: In recent years, many tech companies have rapidly expanded their engineering and produc...
Read more

Apple Battles AI Deepfakes in iOS 27 New Apple Reference Image Feature Authenticates Real Photos.

-
Image
Apple Introduces 'Apple Reference Image' in iOS 27 Beta 5 to Authenticate Genuine Photos Against AI Generation Apple has released iOS 27 Beta 5 to developers, revealing a significant new photographic authentication feature named " Apple Reference Image " First uncovered by 9to5Mac within the Camera app settings, the feature introduces a cryptographic framework designed to verify that a photo was captured using a real iPhone camera rather than generated or altered by artificial intelligence. The feature is disabled by default. When enabled, users can capture photos using a dedicated Reference Mode . Selecting this mode automatically uploads the raw image data and camera metadata to Apple's Private Cloud Compute (PCC) architecture, creating an immutable reference baseline. If the authenticity of an image is later questioned, the photo can be validated against Apple's cloud verification system. Apple emphasizes that privacy remains intact through hardware-bou...
Read more