We're proud to announce the release of hashcat v7.0.0, the result of over two years of development, hundreds of features and fixes, and a complete refactor of several key components. This version also includes all accumulated changes from the v6.2.x minor releases.
This release is huge. If you have 30 minutes, here's the writeup: Full Release Notes and detailed writeup
Here's a quick summary:
- Over 900,000 lines of code changed
- Contributions from 105 developers, including 74 first-time contributors
- Merged and documented all previously unannounced 6.2.x features
 |
Major New Features
- Assimilation Bridge: Integrate external resources like CPUs, FPGAs, embedded interpreters, and more into the cracking pipeline.
- Python Bridge Plugin: Rapidly implement hash-matching logic in Python. No recompilation needed, supports multithreading and rule engine by default.
- Virtual Backend Devices: Internally partitions physical GPUs into multiple logical devices for better bridge integration and async workloads.
- Hash-Mode Autodetection: Omit the
-mflag and let Hashcat detect the hash-mode, or use--identifyto list possibilities. - Docker Build Support: Build Hashcat in a fully containerized, cross-platform environment, including cross-compilation to Windows.
New Algorithm Support
- 58 new application-specific hash types, including Argon2, MetaMask, Microsoft Online Account, SNMPv3, GPG, OpenSSH, and LUKS2
- 17 new generic hash constructions used in real-world web apps and protocols
- 11 new primitives added to the crypto library, improving reuse and plugin development
- 20 new tools to extract hashes from popular sources, including APFS, Virtualbox, BitLocker, and various wallet formats
Performance Improvements
- Complete refactor of the autotuning engine for better device utilization
- Major rewrite of memory management to eliminate previous 4GB allocation caps and enable full memory usage across devices
- Improved tuning for hash-modes like NTLM, NetNTLMv2, and RAR3
- Updated tuning database entries and lower overhead for multi-device setups
- Optimizations to several individual hash-modes including:
- scrypt: up to +320%
- RAR3: up to +54%
- NetNTLMv2: +223% (Intel)
New and Updated Backends
- HIP (AMD): First-class support for AMD's HIP backend, now preferred over OpenCL when both are available
- Metal (Apple): Native GPU support on macOS using Metal, including full Apple Silicon compatibility and major speed improvements
Plugin and Developer Changes
- Improved diagnostics, tokenizer control, and debugging options
- New reusable infrastructure for integrating algorithms directly into both modules and kernels
- Expanded test coverage, edge-case detection, and cross-platform compatibility improvements
Rule Engine Enhancements
- Support for new character class logic and rejection rules, increasing rule engine flexibility
- Refactored and cleaned up rule logic to improve reliability
- Several commonly used rule files have been optimized and expanded