mistral.rs Releases
113 releases of EricLBuehler/mistral.rs
- v0.8.23
Adds Hunyuan v1 model support and introduces ISQ executor/planning, alongside CUDA 13.3 compatibility.
Jun 25, 2026
- v1.38.3
A security-focused release addressing multiple vulnerabilities in the Envoy proxy, including crashes, buffer overflows, and authentication bypasses, while disabling specific extensions.
Jun 23, 2026
- v1.37.5
A security release identical to v1.38.3 in terms of vulnerabilities, resolving the same CVEs and maintaining the disabled DLB extension.
Jun 23, 2026
- v1.36.9
A security release fixing the same set of vulnerabilities as v1.38.3 and v1.37.5, with the same behavioral changes.
Jun 23, 2026
- v1.35.13
A security release addressing a slightly smaller set of vulnerabilities compared to later versions, maintaining the DLB extension disablement.
Jun 23, 2026
- v0.8.22
Minor maintenance release focusing on CUDA release matrix improvements and error handling.
Jun 22, 2026
- 9.4.09.4.0 🥵
A major feature release introducing several new UI components and establishing a sponsorship program to support continued development.
Jun 22, 2026
- v4.15.3
This release focuses on bug fixes for identity and access management, specifically improving token exchange validation, email verification checks, and UI centering for Identity Provider buttons.
Jun 22, 2026
- v0.8.21
Fixes local repo list cache and enhances error messages during model resolution.
Jun 22, 2026
- v0.8.20
Improves CI release workflow reliability and fixes UQFF generation device mapping.
Jun 21, 2026
- v0.8.19
This release focuses on improving sandboxing capabilities by increasing CPU limits, adding support for JSON object response formats, and enhancing the stability of the engine shutdown process.
Jun 21, 2026
- v0.8.18
The update introduces new CLI commands for UQFF handling, improves logging transparency during quantization, and hardens security for remote media fetching operations.
Jun 21, 2026
- v0.8.17
This release adds support for serializing and deserializing unquantized layers and improves the fallback mechanism for AFQ (Automatic Quantization Format).
Jun 20, 2026
- v0.8.16
Distributed training logic is optimized by restricting NCCL usage to multi-node setups, and a bug in the AFQ launch/index calculation for large MoE models is resolved.
Jun 20, 2026
- v0.8.15
This release expands platform support with new CUDA versions and refines agentic workflows by improving tool usage timeouts and lifecycle management.
Jun 19, 2026
- v0.8.14
This release expands platform support with new CUDA versions and refines agentic workflows by improving tool usage timeouts and lifecycle management.
Jun 19, 2026
- v0.8.13
This release expands platform support with new CUDA versions and refines agentic workflows by improving tool usage timeouts and lifecycle management.
Jun 19, 2026
- v0.8.12
The server receives improvements in observability through enhanced logging and Prometheus metrics, alongside new support for Anthropic API skills.
Jun 19, 2026
- v0.8.11
The server receives improvements in observability through enhanced logging and Prometheus metrics, alongside new support for Anthropic API skills.
Jun 19, 2026
- v0.8.10
An agentic feature is added that automatically surfaces files generated during shell or code execution, improving workflow automation.
Jun 18, 2026