Following in the footsteps of competitors Netflix and Disney+, Apple is reportedly working on bringing advertisements to Apple TV+ through an ad-supported tier. MacRumors reports: Apple has apparently been in discussions with the UK's Broadcaster's Audience Research Board (BARB) to explore the necessary data collection techniques for monitoring advertising results. Currently, BARB provides viewing statistics for major UK networks including the BBC, ITV, Channel 4, and Sky, as well as Apple TV+ programming.

While BARB already monitors viewing time for Apple TV+ content, additional techniques are required to track advertising metrics accurately. This data is vital for advertisers to assess the reach and impact of their campaigns on the platform. In addition to the UK, Apple has also reportedly held similar discussions with ratings organizations in the United States. Apple has already included limited advertising in its live sports events, such as last year's Major League Soccer coverage, where ads were incorporated even for Season Pass holders. It is also notable that in March Apple hired Joseph Cady, a former advertising executive from NBCUniversal, to bolster its video advertising team.

Researchers at Check Point have uncovered a clandestine network of approximately 3,000 "ghost" accounts on GitHub, manipulating the platform to promote malicious content. Since June 2023, a cybercriminal dubbed "Stargazer Goblin" has been exploiting GitHub's community features to boost malicious repositories, making them appear legitimate and popular.

Antonis Terefos, a malware reverse engineer at Check Point, discovered the network's activities, which include "starring," "forking," and "watching" malicious pages to increase their visibility and credibility. The network, named "Stargazers Ghost Network," primarily targets Windows users, offering downloads of seemingly legitimate software tools while spreading various types of ransomware and info-stealer malware.

Developer and librarian Misty De Meo, writing about her frustrating experience using GitHub: To me, one of GitHub's killer power user features is its blame view. git blame on the commandline is useful but hard to read; it's not the interface I reach for every day. GitHub's web UI is not only convenient, but the ease by which I can click through to older versions of the blame view on a line by line basis is uniquely powerful. It's one of those features that anchors me to a product: I stopped using offline graphical git clients because it was just that much nicer.

The other day though, I tried to use the blame view on a large file and ran into an issue I don't remember seeing before: I just couldn't find the line of code I was searching for. I threw various keywords from that line into the browser's command+F search box, and nothing came up. I was stumped until a moment later, while I was idly scrolling the page while doing the search again, and it finally found the line I was looking for. I realized what must have happened. I'd heard rumblings that GitHub's in the middle of shipping a frontend rewrite in React, and I realized this must be it. The problem wasn't that the line I wanted wasn't on the page -- it's that the whole document wasn't being rendered at once, so my browser's builtin search bar just couldn't find it. On a hunch, I tried disabling JavaScript entirely in the browser, and suddenly it started working again. GitHub is able to send a fully server-side rendered version of the page, which actually works like it should, but doesn't do so unless JavaScript is completely unavailable.

[...] The corporate branding, the new "AI-powered developer platform" slogan, makes it clear that what I think of as "GitHub" -- the traditional website, what are to me the core features -- simply isn't Microsoft's priority at this point in time. I know many talented people at GitHub who care, but the company's priorities just don't seem to value what I value about the service. This isn't an anti-AI statement so much as a recognition that the tool I still need to use every day is past its prime. Copilot isn't navigating the website for me, replacing my need to the website as it exists today. I've had tools hit this phase of decline and turn it around, but I'm not optimistic. It's still plenty usable now, and probably will be for some years to come, but I'll want to know what other options I have now rather than when things get worse than this.

This week the Computer Science Teachers Association conference kicked off Tuesday in Las Vegas, writes long-time Slashdot reader theodp.

And the "TeachAI" education initiative teamed with the Computer Science Teachers Association to release three briefs "arguing that K-12 computer science education is more important than ever in an age of AI." From the press release: "As AI becomes increasingly present in the classroom, educators are understandably concerned about how it might disrupt the teaching of core CS skills like programming. With these briefs, TeachAI and CSTA hope to reinforce the idea that learning to program is the cornerstone of computational thinking and an important gateway to the problem-solving, critical thinking, and creative thinking skills necessary to thrive in today's digitally driven world. The rise of AI only makes CS education more important."

To help drive home the point to educators, the 39-page Guidance on the Future of Computer Science Education in an Age of AI (penned by five authors from nonprofits CSTA and Code.org) includes a pretty grim comic entitled Learn to Program or Follow Commands. In the panel, two high school students who scoff at the idea of having to learn to code and instead use GenAI to create their Python apps wind up getting stuck in miserable warehouse jobs several years later as a result where they're ordered about by an AI robot.
"The rise of AI only makes CS education more important," according to the group's press release, "with early research showing that people with a greater grasp of underlying computing concepts are able to use AI tools more effectively than those without." A survey by the group also found that 80% of teachers "agree that core concepts in CS education should be updated to emphasize topics that better support learning about AI."

But I'd be curious to hear what Slashdot's readers think. Share your thoughts and opinions in the comments.

Should children still be taught to code in the age of AI?

OpenAI escaped a copyright lawsuit from a group of open-source programmers after they voluntarily dismissed their case against the company in federal court. From a report: The programmers, who allege the generative AI programming tool Copilot was trained on their code without proper attribution, filed their notice of voluntary dismissal Thursday, but will still have their case against GitHub and parent company Microsoft, which collaborated with OpenAI in developing the tool. The proposed class action filed in 2022 in the US District Court for the Northern District of California was the first major copyright case against OpenAI, which has since been hit with numerous lawsuits from authors and news organizations including the New York Times.

The demand for software developers has declined sharply from the peak seen in 2021 and 2022, according to independent analysis by job portal Indeed and research firm ADP, reflecting a broader slowdown in high-paying white-collar job opportunities across tech, marketing, and finance sectors. Nick Bunker, an economist at Indeed, identified these positions as the labor market's current weak point. The shift follows a period of intense recruitment during the pandemic, when tech workers could command premium salaries.

ADP Research adds: Employment of software developers in fact has been slowing since 2020, the year pandemic lockdowns first hit the United States. In January 2024, the U.S. employed fewer software developers than it did six years ago. [...]

The ADP Research Institute tracked employees at 6,500 companies, including more than 75,000 software developers and engineers in 10 industries, between January 2018 and January 2024. Using this data, we built an index to track the employment of software developers beginning in January 2018.

Developer employment grew from January 2018 to November 2019, then began to fall. The index dropped sharply in January 2022 (down 4.6 percentage points), May 2022 (down 3.5 percentage points), and January 2023 (down 3.4 percentage points). Despite intermediate increases in August 2021 and October 2022, the developer employment index has been falling since 2020.

OpenAI on Thursday launched GPT-4o mini, a cost-efficient small AI model that will replace GPT-3.5 Turbo in ChatGPT. Reuters reports: Priced at 15 cents per million input tokens and 60 cents per million output tokens, the GPT-4o mini is more than 60% cheaper than GPT-3.5 Turbo, OpenAI said. It currently outperforms the GPT-4 model on chat preferences and scored 82% on Massive Multitask Language Understanding (MMLU), OpenAI said. MMLU is a textual intelligence and reasoning benchmark used to evaluate the capabilities of language models. A higher MMLU score signifies it can understand and use language better across a variety of domains, enhancing real-world usage.

The GPT-4o mini model's score compared with 77.9% for Google's Gemini Flash and 73.8% for Anthropic's Claude Haiku, according to OpenAI. With the mini model currently supporting text and vision in the application programming interface, OpenAI said support for text, image, video and audio inputs and outputs would be made available in the future.

GitLab, a U.S. provider of cloud-based software development tools whose investors include Google parent Alphabet, is exploring a sale after attracting acquisition interest, Reuters is reporting. From the report: GitLab, which has a market value of about $8 billion, is working with investment bankers on a sale process that has attracted interest from peers, including cloud monitoring firm Datadog, the sources said. Any deal is still weeks away and no agreement is certain, the sources said, requesting anonymity because the matter is confidential.

In its latest State of Application Security Report, Cloudflare says 6.8% of traffic on the internet is malicious, "up a percentage point from last year's study," writes ZDNet's Steven Vaughan-Nichols. "Cloudflare, the content delivery network and security services company, thinks the rise is due to wars and elections. For example, many attacks against Western-interest websites are coming from pro-Russian hacktivist groups such as REvil, KillNet, and Anonymous Sudan." From the report: [...] Distributed Denial of Service (DDoS) attacks continue to be cybercriminals' weapon of choice, making up over 37% of all mitigated traffic. The scale of these attacks is staggering. In the first quarter of 2024 alone, Cloudflare blocked 4.5 million unique DDoS attacks. That total is nearly a third of all the DDoS attacks they mitigated the previous year. But it's not just about the sheer volume of DDoS attacks. The sophistication of these attacks is increasing, too. Last August, Cloudflare mitigated a massive HTTP/2 Rapid Reset DDoS attack that peaked at 201 million requests per second (RPS). That number is three times bigger than any previously observed attack.

The report also highlights the increased importance of application programming interface (API) security. With 60% of dynamic web traffic now API-related, these interfaces are a prime target for attackers. API traffic is growing twice as fast as traditional web traffic. What's worrying is that many organizations appear not to be even aware of a quarter of their API endpoints. Organizations that don't have a tight grip on their internet services or website APIs can't possibly protect themselves from attackers. Evidence suggests the average enterprise application now uses 47 third-party scripts and connects to nearly 50 third-party destinations. Do you know and trust these scripts and connections? You should -- each script of connection is a potential security risk. For instance, the recent Polyfill.io JavaScript incident affected over 380,000 sites.

Finally, about 38% of all HTTP requests processed by Cloudflare are classified as automated bot traffic. Some bots are good and perform a needed service, such as customer service chatbots, or are authorized search engine crawlers. However, as many as 93% of bots are potentially bad.

OW2, the non-profit international consortium dedicated to developing open-source middleware, published an open letter to the European Commission today. They're urging the European Union to continue funding free software after noticing that the Next Generation Internet (NGI) programs were no longer mentioned in Cluster 4 of the 2025 Horizon Europe funding plans.

OW2 argues that discontinuing NGI funding would weaken Europe's technological ecosystem, leaving many projects under-resourced and jeopardizing Europe's position in the global digital landscape. The letter reads, in part: NGI programs have shown their strength and importance to support the European software infrastructure, as a generic funding instrument to fund digital commons and ensure their long-term sustainability. We find this transformation incomprehensible, moreover when NGI has proven efficient and economical to support free software as a whole, from the smallest to the most established initiatives. This ecosystem diversity backs the strength of European technological innovation, and maintaining the NGI initiative to provide structural support to software projects at the heart of worldwide innovation is key to enforce the sovereignty of a European infrastructure. Contrary to common perception, technical innovations often originate from European rather than North American programming communities, and are mostly initiated by small-scaled organizations.

Previous Cluster 4 allocated 27 millions euros to:
- "Human centric Internet aligned with values and principles commonly shared in Europe";
- "A flourishing internet, based on common building blocks created within NGI, that enables better control of our digital life";
- "A structured eco-system of talented contributors driving the creation of new internet commons and the evolution of existing internet commons."

In the name of these challenges, more than 500 projects received NGI funding in the first 5 years, backed by 18 organizations managing these European funding consortia.

Thunderbird's annual Extended Support Release was revealed Friday, promising "significant" improvements to the overall user experience and "the speed at which we can deliver new features to you," according to the Thunderbird blog: We've devoted significant development time integrating Rust — a modern programming language originally created by Mozilla Research — into Thunderbird. Even though this is a seemingly invisible change, it is a major leap forward because it enhances our code quality and performance. This overhaul will allow us to share features between the desktop and future mobile versions of Thunderbird, and speed up our development process. It's a win for our developers and a win for you.
More from the blog OMG Ubuntu: I'm also stoked to see that Thunderbird 128 makes 'newest first' the default sort order for messages in message list. While some prefer the old way, I always found it strange that the oldest mails were shown first — team reverse chronology, represent!
They also cite "a number of OpenPGP improvements," plus a new preference option for displaying full names and email addresses of all recipients in the message list. (Plus, threaded-message views now display a "New Message" count.)

Other new features in this release:

• A new and more attractive layout for Cards View (with adjustable heights) that "makes it easier to scan your email threads and glean information."

• The folder pane has better recall of message thread states

• Improved theme compatibility. "Your Thunderbird should blend seamlessly with your desktop environment, matching the system's accent colors perfectly." (Especially beneficial on Ubuntu and Mint.)

• You can now customize the color of your account icon.

The Thunderbird blog also mentions that "We plan to launch the first phase of built-in support for Exchange, as well as Mozilla Sync, in a future Nebula point release (e.g. Thunderbird 128.X)."

An anonymous reader shared this report from InfoWorld: Rust has leaped to its highest position ever in the monthly Tiobe index of language popularity, scaling to the 13th spot this month, with placement in the top 10 anticipated in an upcoming edition. Previously, Rust has never gone higher than 17th place in the Tiobe Programming Index. Tiobe CEO Paul Jansen attributed Rust's ascent in the just-released July index to a February 2024 U.S. White House report recommending Rust over C/C+ for safety reasons. He also credited the growing community and ecosystem support for the language. "Rust is finally moving up."
The article adds that these rankings are based on "the number of skilled engineers worldwide, courses, and third-party vendors pertaining to languages, examining websites such as Google, Amazon, Wikipedia, and more than 20 others to determine the monthly numbers."

1. Python
2. C++
3. C
4. Java
5. C#
6. JavaScript
7. Go
8. Visual Basic
9. Fortran
10. SQL

Interestingly, Rust has just moved into the top ten on the rival rankings from the rival Pypl Popularity of Programming Language index (which according to the article "assesses how often languages are searched on in Google.")

1. Python
2. Java
3. JavaScript
4. C#
5. C/C++
6. R
7. PHP
8. TypeScript
9. Swift
10. Rust

snydeq shares a report from CSO Online, written by Lucian Constantin: A personal GitHub access token with administrative privileges to the official repositories for the Python programming language and the Python Package Index (PyPI) was exposed for over a year. The access token belonged to the Python Software Foundation's director of infrastructure and was accidentally included in a compiled binary file that was published as part of a container image on Docker Hub. [...] The incident shows that scrubbing access tokens from source code only, which some development tools do automatically, is not enough to prevent potential security breaches. Sensitive credentials can also be included in environment variables, configuration files and even binary artifacts as a result of automated build processes and developer mistakes. "Although we encounter many secrets that are leaked in the same manner, this case was exceptional because it is difficult to overestimate the potential consequences if it had fallen into the wrong hands -- one could supposedly inject malicious code into PyPI packages (imagine replacing all Python packages with malicious ones), and even to the Python language itself," researchers from security firm JFrog, who found and reported the token, wrote in a report.

An anonymous reader shared this report from NPR: A teenage computer whiz who used the early-aughts internet to spread awareness of the Catholic faith will become the church's first millennial saint.

Carlo Acutis, who died of leukemia at age 15 in 2006, is already referred to as "God's influencer" and the "patron saint of the internet" for his work cataloging Eucharistic miracles around the world — and soon it will be official. Pope Francis and a group of cardinals approved Acutis for canonization at a meeting at the Vatican on Monday, Vatican News announced. It says he will likely be proclaimed a saint at some point in 2025, during the church's jubilee year.

Acutis was a devout Catholic who taught himself programming from an early age and created websites with a spiritual focus, including his widely praised database of miracles. He is credited with helping homeless people and defending victims of bullying during his lifetime, and having a hand in two healing miracles after his death — the requisite number for all Catholic saints. Monday's approval clears the final hurdle in a multiyear process, which began in 2013 when the pope approved the cause for his beatification and canonization and named him "a Servant of God...."

Acutis also loved playing video games — CNN cited Halo, Super Mario and Pokémon among his favorites — though limited himself to one hour a week.

IEEE Spectrum (the IEEE's official publication) asks the question. "How does an AI code generator compare to a human programmer?" A study published in the June issue of IEEE Transactions on Software Engineering evaluated the code produced by OpenAI's ChatGPT in terms of functionality, complexity and security. The results show that ChatGPT has an extremely broad range of success when it comes to producing functional code — with a success rate ranging from anywhere as poor as 0.66 percent and as good as 89 percent — depending on the difficulty of the task, the programming language, and a number of other factors. While in some cases the AI generator could produce better code than humans, the analysis also reveals some security concerns with AI-generated code.
The study tested GPT-3.5 on 728 coding problems from the LeetCode testing platform — and in five programming languages: C, C++, Java, JavaScript, and Python. The results? Overall, ChatGPT was fairly good at solving problems in the different coding languages — but especially when attempting to solve coding problems that existed on LeetCode before 2021. For instance, it was able to produce functional code for easy, medium, and hard problems with success rates of about 89, 71, and 40 percent, respectively. "However, when it comes to the algorithm problems after 2021, ChatGPT's ability to generate functionally correct code is affected. It sometimes fails to understand the meaning of questions, even for easy level problems," said Yutian Tang, a lecturer at the University of Glasgow. For example, ChatGPT's ability to produce functional code for "easy" coding problems dropped from 89 percent to 52 percent after 2021. And its ability to generate functional code for "hard" problems dropped from 40 percent to 0.66 percent after this time as well...

The researchers also explored the ability of ChatGPT to fix its own coding errors after receiving feedback from LeetCode. They randomly selected 50 coding scenarios where ChatGPT initially generated incorrect coding, either because it didn't understand the content or problem at hand. While ChatGPT was good at fixing compiling errors, it generally was not good at correcting its own mistakes... The researchers also found that ChatGPT-generated code did have a fair amount of vulnerabilities, such as a missing null test, but many of these were easily fixable.
"Interestingly, ChatGPT is able to generate code with smaller runtime and memory overheads than at least 50 percent of human solutions to the same LeetCode problems..."

"After approximately seven years in development, the Eclipse Foundation's Theia IDE project is now generally available," writes ADT magazine, "emerging from beta to challenge Microsoft's similar Visual Studio Code (VS Code) editor." The Eclipse Theia IDE is part of the Eclipse Cloud DevTools ecosystem. The Eclipse Foundation calls it "a true open-source alternative to VS Code," which was built on open source but includes proprietary elements, such as default telemetry, which collects usage data...

Theia was built on the same Monaco editor that powers VS Code, and it supports the same Language Server Protocol (LSP) and Debug Adapter Protocol (DAP) that provide IntelliSense code completions, error checking and other features. The Theia IDE also supports the same extensions as VS Code (via the Open VSX Registry instead of Microsoft's Visual Studio Code Marketplace), which are typically written in TypeScript and JavaScript. There are many, many more extensions available for VS Code in Microsoft's marketplace, while "Extensions for VS Code Compatible Editors" in the Open VSX Registry number 3,784 at the time of this writing...

The Eclipse Foundation emphasized another difference between its Theia IDE and VS Code: the surrounding ecosystem/community. "At the core of Theia IDE is its vibrant open source community hosted by the Eclipse Foundation," the organization said in a news release. "This ensures freedom for commercial use without proprietary constraints and fosters innovation and reliability through contributions from companies such as Ericsson, EclipseSource, STMicroelectronics, TypeFox, and more. The community-driven model encourages participation and adaptation according to user needs and feedback."

Indeed, the list of contributors to and adopters of the platform is extensive, also featuring Broadcom, Arm, IBM, Red Hat, SAP, Samsung, Google, Gitpod, Huawei and many others.
The It's FOSS blog has some screenshots and a detailed rundown.

ADT magazine stresses that there's also an entirely distinct (but related) project called the Eclipse Theia Platform (not IDE) which differs from VS Code by allowing developers "to create desktop and cloud IDEs using a single, open-source technology stack" [that can be used in open-source initiatives]. The Eclipse Theia platform "allows developers to customize every aspect of the IDE without forking or patching the code... fully tailored for the needs of internal company projects or for commercial resale as a branded product."

USENIX, the long-running OS/networking research group, also publishes a magazine called ;login:. Today the magazine's editor — security consultant Rik Farrow — stopped by Slashdot to share some new research. rikfarrow writes: Caching means using faster memory to store frequently requested data, and the most commonly used algorithm for determining which items to discard when the cache is full is Least Recently Used [or "LRU"]. These researchers have come up with a more efficient and scalable method that uses just a few lines of code to convert LRU to SIEVE.
Just like a sieve, it sifts through objects (using a pointer called a "hand") to "filter out unpopular objects and retain the popular ones," with popularity based on a single bit that tracks whether a cached object has been visited: As the "hand" moves from the tail (the oldest object) to the head (the newest object), objects that have not been visited are evicted... During the subsequent rounds of sifting, if objects that survived previous rounds remain popular, they will stay in the cache. In such a case, since most old objects are not evicted, the eviction hand quickly moves past the old popular objects to the queue positions close to the head. This allows newly inserted objects to be quickly assessed and evicted, putting greater eviction pressure on unpopular items (such as "one-hit wonders") than LRU-based eviction algorithms.
It's an example of "lazy promotion and quick demotion". Popular objects get retained with minimal effort, with quick demotion "critical because most objects are not reused before eviction."

After 1559 traces (of 247,017 million requests to 14,852 million objects), they found SIEVE reduces the miss ratio (when needed data isn't in the cache) by more than 42% on 10% of the traces with a mean of 21%, when compared to FIFO. (And it was also faster and more scalable than LRU.)

"SIEVE not only achieves better efficiency, higher throughput, and better scalability, but it is also very simple."

Michigan's House of Representatives passed a bill requiring all the state's public high schools to offer a computer science course by the start of the 2027-28 school year. (The bill now goes to the Senate, according to a report from Chalkbeat Detroit.)

Long-time Slashdot reader theodp writes: Michigan is also removing the requirement for CS teacher endorsements in 2026, paving the way for CS courses to be taught in 2027 by teachers who have "demonstrated strong computer science skills" but do not hold a CS endorsement. Michigan's easing of CS teaching requirements comes in the same year that New York State will begin requiring credentials for all CS teachers.

With lobbyist Julia Wynn from the tech giant-backed nonprofit Code.org sitting at her side, Michigan State Rep. Carol Glavnille introduced the CS bill (HB5649) to the House in May (hearing video, 16:20). "This is not a graduation requirement," Glavnille emphasized in her testimony. Code.org's Wynn called the Bill "an important first step" — after all, Code.org's goal is "to require all students to take CS to earn a HS diploma" — noting that Code.org has also been closely collaborating with Michigan's Education department "on the language and the Bill since inception." Wynn went on to inform lawmakers that "even just attending a high school that offers computer science delivers concrete employment and earnings benefits for students," citing a recent Brookings Institute article that also noted "30 states have adopted a key part of Code.org Advocacy Coalition's policy recommendations, which require all high schools to offer CS coursework, while eight states (and counting) have gone a step further in requiring all students to take CS as a high school graduation requirement."

Minutes from the hearing report other parties submitting cards in support of HB 5649 included Amazon (a $3+ million Code.org Platinum Supporter) and AWS (a Code.org In-Kind Supporter), as well as College Board (which offers the AP CS A and CSP exams) and TechNet (which notes its "teams at the federal and state levels advocate with policymakers on behalf of our member companies").

The Ontario Science Center, a world-class science and cultural institution in Toronto, is shutting down immediately due to the risk that the building's roof could collapse, the province announced Friday. CBC News: The abrupt closure, which the province says could last years, comes after the government's controversial announcement in 2023 that the popular landmark and attraction would be moved to the Ontario Place site -- a move it says will save costs. "The actions taken today will protect the health and safety of visitors and staff," said Infrastructure Minister Kinga Surma in a news release. "We are making every effort to avoid disruption to the public and help the Ontario Science Centre continue delivering on its mandate."

An engineering report this week by Rimkus Consulting Group showed each of the centre's three buildings contain roof panels in a "distressed, high-risk" condition, the Ministry of Infrastructure said in a news release. The panels require fixing by Oct. 31, 2024 to "avoid further stress due to potential snow load which could lead to roof panel failure," the release said. Fixing the roof will cost between $22 million and $40 million, the ministry said, requiring the centre be closed for up to two years. "These estimates are incomplete and subject to change," said the ministry, noting the costs make up only a "small portion" of the funding needed to keep the science centre open. The government says the centre needs $478 million to tackle its "failing infrastructure" and sustain programming.

Shubham Sharma reports via VentureBeat: Chinese AI startup DeepSeek, which previously made headlines with a ChatGPT competitor trained on 2 trillion English and Chinese tokens, has announced the release of DeepSeek Coder V2, an open-source mixture of experts (MoE) code language model. Built upon DeepSeek-V2, an MoE model that debuted last month, DeepSeek Coder V2 excels at both coding and math tasks. It supports more than 300 programming languages and outperforms state-of-the-art closed-source models, including GPT-4 Turbo, Claude 3 Opus and Gemini 1.5 Pro. The company claims this is the first time an open model has achieved this feat, sitting way ahead of Llama 3-70B and other models in the category. It also notes that DeepSeek Coder V2 maintains comparable performance in terms of general reasoning and language capabilities.

Founded last year with a mission to "unravel the mystery of AGI with curiosity," DeepSeek has been a notable Chinese player in the AI race, joining the likes of Qwen, 01.AI and Baidu. In fact, within a year of its launch, the company has already open-sourced a bunch of models, including the DeepSeek Coder family. The original DeepSeek Coder, with up to 33 billion parameters, did decently on benchmarks with capabilities like project-level code completion and infilling, but only supported 86 programming languages and a context window of 16K. The new V2 offering builds on that work, expanding language support to 338 and context window to 128K -- enabling it to handle more complex and extensive coding tasks. When tested on MBPP+, HumanEval, and Aider benchmarks, designed to evaluate code generation, editing and problem-solving capabilities of LLMs, DeepSeek Coder V2 scored 76.2, 90.2, and 73.7, respectively -- sitting ahead of most closed and open-source models, including GPT-4 Turbo, Claude 3 Opus, Gemini 1.5 Pro, Codestral and Llama-3 70B. Similar performance was seen across benchmarks designed to assess the model's mathematical capabilities (MATH and GSM8K). The only model that managed to outperform DeepSeek's offering across multiple benchmarks was GPT-4o, which obtained marginally higher scores in HumanEval, LiveCode Bench, MATH and GSM8K. [...]

As of now, DeepSeek Coder V2 is being offered under a MIT license, which allows for both research and unrestricted commercial use. Users can download both 16B and 236B sizes in instruct and base avatars via Hugging Face. Alternatively, the company is also providing access to the models via API through its platform under a pay-as-you-go model. For those who want to test out the capabilities of the models first, the company is offering the option to interact. with Deepseek Coder V2 via chatbot.