GPT‑5.6 is priced per 1M tokens across three model sizes: Sol is $5 input / $30 output; Terra is $2.50 input / $15 output; and Luna is $1 input / $6 output. GPT‑5.6 also introduces more predictable prompt caching, including support for explicit cache breakpoints and a 30-minute minimum cache life. For GPT‑5.6 and later models, cache writes are billed at 1.25x the model’s uncached input rate, while cache reads continue to receive the 90% cached-input discount.

We're also launching GPT‑5.6 Sol on Cerebras at up to 750 tokens per second in July, bringing frontier intelligence to customers at unprecedented speed. Access will initially be limited to select customers as we expand capacity.

Nuitka is an optimizing Python compiler that translates Python scripts into optimized C or C++ code. It then builds this into standalone, executable binaries (like .exe files) that run independently, without requiring the user to have Python installed.

A GitHub account is mass-dropping exploit PoCs framed as undisclosed 0-days, with a note telling readers to report them and "take credit for the CVE" themselves.

Coordinated disclosure, minus the coordination.

libssh2 through 1.11.1, fixed in commit 97acf3d contains an out-of-bounds write vulnerability in ssh2_transport_read() that fails to enforce upper bounds on packet_length field. Remote attackers can send crafted SSH packets with excessively large packet_length values to corrupt heap memory and achieve remote code execution.

A malicious OpenVPN server can push an echo option that decodes into script.win.user.disconnect. OpenVPN Connect later executes that command on disconnect even though the imported profile's script permission state remains unset or false.

Verified impact:
- Current-user arbitrary command execution on VPN disconnect.
- Import alone is not enough. The client must connect, receive the pushed echo value, and then disconnect.
- The default payload writes %TEMP%\openvpn_connect_echo_script_ace_marker.txt.

A malicious OpenVPN server can push dhcp-option PROXY_AUTO_CONFIG_URL. OpenVPN Connect passes the pushed PAC URL through the privileged /tun-setup path, and the LocalSystem agent applies the proxy action by impersonating the current user. During the VPN session, HKCU Internet Settings receives the server-controlled AutoConfigURL; OpenVPN Connect clears it on disconnect.

It applies a range of AST-based transformations to make source code significantly harder to read, analyze, and reverse engineer.
These include techniques such as control-flow flattening, constant encryption, and other Lua-specific obfuscation strategies.
The project was inspired by the excellent javascript-obfuscator.

GNU gzip contains a global buffer overflow vulnerability in the LZH decompression logic caused by improper reuse of shared global state between different decompression formats within a single execution. GNU gzip maintains a global array that is shared across the LZ77, LZW, and LZH decompression routines and is not reinitialized between files processed in the same invocation.
By decompressing a specially crafted LZW file followed by a specially crafted LZH file in a single gzip -d command, an attacker can poison the shared global state and subsequently trigger an out‑of‑bounds read in the LZH decoder. The LZH decompression logic follows stale values left in the shared array, causing reads past the end of the allocated global buffer.

This issue has been fixed in the commit 63dbf6b3b9e6e781df1a6a64e609b10e23969681

GNU gzip contains a vulnerability in the gzexe utility related to insecure temporary file handling. When the mktemp utility is not available in the user’s PATH, gzexe falls back to constructing a temporary file path based solely on the process ID (PID). This predictable filename is created without exclusive access or existence checks.
A local attacker can pre‑create the predicted temporary file path as a symbolic link pointing to an arbitrary file writable by the victim. When gzexe runs, it follows the symlink and overwrites the target file, resulting in a time‑of‑check to time‑of‑use (TOCTOU) condition that allows arbitrary file overwrite.

This issue has been fixed in the commit 4e6f8b24ab823146ab8776f0b7fe486ab34d4269

Prior to deno v2.8.1, environment access is gated by the env permission. You can deny it with --deny-env, or restrict it to a specific allowlist with --allow-env=FOO,BAR. The expectation is that a program running without env permission cannot change process.env. process.loadEnvFile() (the Node-compatible API for loading variables from a .env file) does not honor this. It only checks that the program has read permission for the dotenv file, then writes every key in that file into the process environment, even when env access is denied.
In effect, --allow-read plus a writable or attacker-controlled .env file is enough to defeat --deny-env. This vulnerability is fixed in v2.8.1.

Prior to pnpm v10.33.4 and v11.0.7, a malicious codeload.github.com server can serve whatever tarball it wants and pnpm will install it regardless of the lockfile. The lockfile does not store the hash of the dependencies from https://codeload.github.com. This means that if this server was compromised or a person's machine configuration was compromised, pnpm would download and install these dependencies. This vulnerability is fixed in v10.33.4 and v11.0.7.

A missing authorization vulnerability was identified in GitHub Enterprise Server that allowed an authenticated user to read source code from private repositories they did not have access to. The
Copilot pull request description diff summary endpoint accepted a cross-repository comparison range and rendered the resulting diff without verifying that the requesting user was authorized to view
the target repository. Exploitation required an authenticated account on the instance with read access to at least one repository to use as the comparison base. This vulnerability affected all
versions of GitHub Enterprise Server prior to 3.21 and was fixed in versions 3.17.17, 3.18.11, 3.19.8, and 3.20.4. This vulnerability was reported via the GitHub Bug Bounty program.

CVE-2026-14432
Use after free in V8 in Google Chrome prior to 150.0.7871.46 allowed a remote attacker to execute arbitrary code inside a sandbox via a crafted HTML page. (Chromium security severity: Medium)

CVE-2026-14424
Use after free in Dawn in Google Chrome on Mac prior to 150.0.7871.46 allowed a remote attacker to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High)

CVE-2026-14427, CVE-2026-14429, CVE-2026-14419
Insufficient validation of untrusted input in Skia in Google Chrome prior to 150.0.7871.46 allowed a remote attacker who had compromised the renderer process to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High/Critical)

CVE-2026-14428
Insufficient validation of untrusted input in Dawn in Google Chrome on Android prior to 150.0.7871.46 allowed a remote attacker who had compromised the renderer process to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High)

CVE-2026-14420
Out of bounds read and write in Dawn in Google Chrome prior to 150.0.7871.46 allowed a remote attacker to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: Critical)

CVE-2026-14425, CVE-2026-14411
Use after free in ANGLE in Google Chrome prior to 150.0.7871.46 allowed a remote attacker to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High)

CVE-2026-14423
Type Confusion in Tint in Google Chrome prior to 150.0.7871.46 allowed a remote attacker to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High)

A critical arbitrary code execution vulnerability affecting KDE Plasma was disclosed, allowing sandbox escapes via the "Open New Window" action. The exploit, discovered by Kimiblock, lets malicious sandboxed apps execute arbitrary code on the host system. The issue was reported to KDE but remains unpatched, leading to public disclosure after the 90-day embargo. More details and a proof-of-concept are available on the blog.