Mistral: Devstral Small 2505

Text input Text output Free Option
Author's Description

Devstral-Small-2505 is a 24B parameter agentic LLM fine-tuned from Mistral-Small-3.1, jointly developed by Mistral AI and All Hands AI for advanced software engineering tasks. It is optimized for codebase exploration, multi-file editing, and integration into coding agents, achieving state-of-the-art results on SWE-Bench Verified (46.8%). Devstral supports a 128k context window and uses a custom Tekken tokenizer. It is text-only, with the vision encoder removed, and is suitable for local deployment on high-end consumer hardware (e.g., RTX 4090, 32GB RAM Macs). Devstral is best used in agentic workflows via the OpenHands scaffold and is compatible with inference frameworks like vLLM, Transformers, and Ollama. It is released under the Apache 2.0 license.

Key Specifications
Cost
$
Context
128K
Parameters
24B (Rumoured)
Released
May 21, 2025
Speed
Ability
Reliability
Supported Parameters

This model supports the following parameters:

Response Format Stop Seed Min P Top P Max Tokens Frequency Penalty Temperature Presence Penalty
Features

This model supports the following features:

Response Format
Performance Summary

Mistral: Devstral Small 2505 demonstrates a strong overall performance profile, particularly excelling in reliability and cost-efficiency. It consistently achieves a 100% success rate across all benchmarks, indicating exceptional stability and minimal technical failures. The model also offers highly competitive pricing, ranking in the 88th percentile. While its speed performance is generally good, placing in the 67th percentile, it shows some variability across tasks. In terms of specific capabilities, Devstral Small 2505 exhibits perfect accuracy in one of the Instruction Following benchmarks, highlighting its precision in executing complex directives. It also performs well in General Knowledge (98.0%) and Ethics (98.0%). Its specialized fine-tuning for software engineering tasks is reflected in a solid 82.0% accuracy in Coding. However, its performance in Hallucinations (94.0%) and Mathematics (84.0%) is around the 50th percentile, suggesting areas for potential improvement. The model's 49.0% accuracy in a second Instruction Following benchmark indicates a potential inconsistency or sensitivity to specific instruction complexities.

Model Pricing

Current Pricing

Feature Price (per 1M tokens)
Prompt $0.06
Completion $0.12

Price History

Available Endpoints
Provider Endpoint Name Context Length Pricing (Input) Pricing (Output)
DeepInfra
DeepInfra | mistralai/devstral-small-2505 128K $0.06 / 1M tokens $0.12 / 1M tokens
NextBit
NextBit | mistralai/devstral-small-2505 131K $0.04 / 1M tokens $0.14 / 1M tokens
Parasail
Parasail | mistralai/devstral-small-2505 131K $0.04 / 1M tokens $0.14 / 1M tokens
Mistral
Mistral | mistralai/devstral-small-2505 131K $0.1 / 1M tokens $0.3 / 1M tokens
Nebius
Nebius | mistralai/devstral-small-2505 128K $0.08 / 1M tokens $0.24 / 1M tokens
Chutes
Chutes | mistralai/devstral-small-2505 131K $0.04 / 1M tokens $0.14 / 1M tokens
Chutes
Chutes | mistralai/devstral-small-2505 131K $0.04 / 1M tokens $0.14 / 1M tokens
Benchmark Results
Benchmark Category Reasoning Strategy Free Executions Accuracy Cost Duration
Other Models by mistralai