Mistral launches Devstral 2, a 24B model with laptop version for code tasks
Mistral AI’s latest release, Devstral 2, pushes the envelope for on‑device programming assistance. The new 24‑billion‑parameter model arrives with a stripped‑down variant that runs on a typical laptop, promising developers the ability to run sophisticated code‑generation workloads without cloud dependence. This follows a previous rollout that paired Mistral with All Hands AI and shipped under an Apache 2.0 licence, a move that signaled the company’s commitment to open‑source tooling for autonomous coding tasks.
By offering a version that fits on consumer hardware, Mistral aims to broaden access beyond large‑scale data‑center users, addressing a niche where developers need quick, locally‑executed suggestions for refactoring, file traversal, and multi‑step reasoning. The shift from a purely server‑based offering to a hybrid approach raises questions about performance trade‑offs and the practical limits of on‑device inference. It also sets the stage for a deeper look at how the company has evolved its “agentic” code model since its initial launch.
One year later, the company followed up with Devstral, a 24B model purpose‑built for "agentic" behavior—handling long‑range reasoning, file navigation, and autonomous code modification. Released in partnership with All Hands AI and licensed under Apache 2.0, Devstral was notable not just for its po
One year later, the company followed up with Devstral, a 24B model purpose-built for "agentic" behavior--handling long-range reasoning, file navigation, and autonomous code modification. Released in partnership with All Hands AI and licensed under Apache 2.0, Devstral was notable not just for its portability (it could run on a MacBook or RTX 4090), but for its performance: it beat out several closed models on SWE-Bench Verified, a benchmark of 500 real-world GitHub issues. Then came Mistral 3, announced in December 2025 as a portfolio of 10 open-weight models targeting everything from drones and smartphones to cloud infrastructure.
Will developers adopt the laptop‑friendly variant? Mistral’s Devstral 2 arrives as a 24‑billion‑parameter model paired with a stripped‑down version that can run on consumer hardware. The company’s recent rollout follows the open‑source Mistral 3 family, which targeted edge devices and local deployments.
By licensing Devstral 2 under Apache 2.0 and collaborating with All Hands AI, Mistral continues its open‑source strategy. The new pair of models is tuned for software‑engineering tasks, promising long‑range reasoning and autonomous code modification, capabilities highlighted in the earlier Devstral release. Yet, it is unclear whether the smaller footprint will translate into practical productivity gains for indie developers or enterprise teams.
The startup has emerged from a year of public questioning, offering tools aimed at both markets, but adoption metrics remain unknown. As the December 2025 launch demonstrates technical ambition, the broader community’s response will determine whether the models fulfill their intended role. Stakeholders will likely monitor performance benchmarks and real‑world case studies before integrating the models into production pipelines.
Without independent evaluations, confidence in the claimed capabilities stays tentative.
Further Reading
- Introducing: Devstral 2 and Mistral Vibe CLI – State-of-the-art, open-source agentic coding models and terminal AI - Mistral AI
- Mistral AI surfs vibe-coding tailwinds with new coding models - TechCrunch
- Mistral’s “Open Source” Trick: Build a Great Model, Gate It Behind Revenue Caps, Call It Freedom - Implicator
- Latest news: Introducing Devstral 2 and Mistral Vibe CLI - Mistral AI
More in Open Source
HBO Max launches 12 themed channels streaming Friends, Game of Thrones 24/7
HBO Max is rolling out a dozen dedicated streams that run marquee titles—think Friends reruns and...
Author defends USD 40 cable paperweight over cheap clips for kitchen table
When a modest‑priced gadget becomes the centerpiece of a kitchen work surface, the debate isn’t...
Google adds Gemini AI tools to Chrome for iPhone and iPad users
Why does this matter for the millions of iOS and iPadOS users who rely on Chrome as their default...
Nous Research's Nomos 1 ranks second on Putnam, trailing DeepSeekMath-V2
Nous Research just dropped Nomos 1, an open‑source model that managed to clinch second place on the...
Mistral AI launches DevStral 2 (123B), DevStral Small 2 (24B) and Vibe CLI
Why does this matter now? Open‑source AI developers have been waiting for a model that can handle...
Common Questions Answered
What is the size and licensing of Mistral's Devstral 2 model?
Devstral 2 is a 24‑billion‑parameter model that is released under the Apache 2.0 open‑source license. This licensing choice allows developers to freely use, modify, and distribute the model for their own code‑generation workloads.
How does the stripped‑down variant of Devstral 2 enable laptop‑friendly code generation?
The stripped‑down variant is optimized to run on typical consumer hardware such as a standard laptop, eliminating the need for cloud resources. By reducing memory and compute requirements, it still supports sophisticated software‑engineering tasks while remaining portable.
What benchmark did the original Devstral model excel on, and what does that indicate?
The original Devstral model outperformed several closed‑source competitors on the SWE‑Bench Verified benchmark, which consists of 500 real‑world GitHub issues. This result demonstrates its strong capability in handling realistic code‑generation and debugging scenarios.
In what way does Devstral 2 build on the earlier Mistral 3 family’s focus on edge devices?
Devstral 2 continues the Mistral 3 family’s emphasis on edge deployment by offering a version that can run locally on consumer devices, extending the open‑source strategy to software‑engineering workloads. The model’s tuning for code‑related tasks makes it suitable for on‑device development environments.