Devstral 2 logo
Mistral
•

Open weights model

•

Released December 2025

Devstral 2: Analyse von Intelligenz, Leistung und Preis

Ausprobieren Benchmarks der API-Anbieter

Modellübersicht

Intelligenz

19
Artificial Analysis Intelligence Index
4 von 4 Einheiten für Intelligenz.

Geschwindigkeit

42.8
Ausgabetokens pro Sekunde
1 von 4 Einheiten für Geschwindigkeit.

Eingabepreis

0,00 $
USD pro 1 Mio. Tokens
1 von 4 Einheiten für Eingabepreis.

Ausgabepreis

0,00 $
USD pro 1 Mio. Tokens
1 von 4 Einheiten für Ausgabepreis.

Ausführlichkeit

9.2M
Ausgabetokens im Intelligence Index
2 von 4 Einheiten für Ausführlichkeit.

Devstral 2 zählt bei der Intelligenz zu den führenden Modellen und ist preislich gut aufgestellt im Vergleich zu anderen Modellen ohne Reasoning mit offenen Gewichten ähnlicher Größe. Das Modell ist außerdem auffallend langsam, antwortet jedoch recht knapp. Das Modell unterstützt Text als Eingabe und Text als Ausgabe. Das Kontextfenster umfasst 256k Tokens.

Devstral 2 erzielt im Artificial Analysis Intelligence Index einen Wert von 19 und liegt damit deutlich über dem Durchschnitt der vergleichbaren Modelle (Median: 7). Bei der Evaluation des Intelligence Index erzeugte es 9.2M Tokens und antwortete damit recht knapp im Vergleich zum Median von 14M.

Der Preis für Devstral 2 beträgt $0.00 pro 1 Mio. Eingabetokens (preislich konkurrenzfähig; Median: $0.18) und $0.00 pro 1 Mio. Ausgabetokens (preislich konkurrenzfähig; Median: $0.40). Insgesamt kostete die Evaluation von Devstral 2 im Intelligence Index $0.00.

Mit 43 Tokens pro Sekunde ist Devstral 2 auffallend langsam (Median: 96).

ReasoningNein

Diese Seite zeigt die Version dieses Modells ohne Reasoning.

Möglicherweise gibt es auch eine Reasoning-Variante.

Eingabemodalität

Unterstützt: Text

Ausgabemodalität

Unterstützt: Text

Kontextfenster256k
~384 A4-Seiten in Arial mit Schriftgröße 12
Gesamtparameter125B
LizenzModified MIT License
ModellgewichteHugging Face

Metriken werden mit Modellen derselben Klasse verglichen:

  • Modelle ohne Reasoning → Vergleich nur mit anderen Modellen ohne Reasoning
  • Reasoning-Modelle → Vergleich mit Reasoning-Modellen und Modellen ohne Reasoning
  • Modelle mit offenen Gewichten → Vergleich nur mit anderen Modellen mit offenen Gewichten derselben Größenklasse:
    • Sehr klein: ≤4B Parameter
    • Klein: 4B–40B Parameter
    • Mittel: 40B–150B Parameter
    • Groß: >150B Parameter
  • Proprietäre Modelle → Vergleich mit proprietären Modellen und Modellen mit offenen Gewichten derselben Preisklasse anhand eines Mischpreisverhältnisses von 3:1 für Eingabe/Ausgabe:
    • <$0.15 pro 1 Mio. Tokens
    • $0.15–$1 pro 1 Mio. Tokens
    • >$1 pro 1 Mio. Tokens

Wichtigste Ergebnisse

Intelligenz

Artificial Analysis Intelligence Index · Higher is better

Geschwindigkeit

Output tokens per second · Higher is better

Kosten pro Aufgabe

Weighted average cost (USD) per Intelligence Index task · Lower is better

Intelligenz

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Artificial Analysis Intelligence Index by Open Weights / Proprietary

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Indicates whether the model weights are available. Models are labelled as 'Commercial Use Restricted' if the weights are available but commercial use is limited (typically requires obtaining a paid license).

Benchmarks

Intelligence Evaluations

Intelligence evaluations measured independently by Artificial Analysis · Higher is better
GDPval-AA v2

Agentic real-world work tasks, (Elo-500)/2000

𝜏³-Banking

Agentic tool use

Terminal-Bench v2.1

Agentic coding & terminal use

SciCode

Coding

Humanity's Last Exam

Reasoning & knowledge

GPQA Diamond

Scientific reasoning

CritPt

Physics reasoning

AA-Omniscience Accuracy

Knowledge

AA-Omniscience Non-Hallucination Rate

1 - hallucination rate

AA-LCR

Long context reasoning

AA-Briefcase

Agentic knowledge work, Elo

AutomationBench-AA

Agentic SaaS workflows

Harvey LAB-AA

Legal agentic work, criterion pass rate

EnterpriseOps-Gym-AANew

Agentic business operations

IFBench

Instruction following

APEX-Agents-AA

Long-horizon agentic tasks

ITBench-AA

Kubernetes incident root-cause analysis

MMMU-Pro

Visual reasoning

Reasoning models are indicated by a lightbulb icon

While model intelligence generally translates across use cases, specific evaluations may be more relevant for certain use cases.

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

AA-Omniscience

AA-Omniscience Index

AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.
Reasoning models are indicated by a lightbulb icon

AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.

Openness Index

Artificial Analysis Openness Index: Score

Openness Index assesses model openness on a 0 to 100 normalized scale (higher is more open)
Reasoning models are indicated by a lightbulb icon

Vergleiche des Intelligence Index

Intelligence Index vs. Cost per Intelligence Index Task

Artificial Analysis Intelligence Index · Weighted average cost (USD) per Artificial Analysis Intelligence Index task
Most attractive quadrant
Pareto line
Reasoning models are indicated by a lightbulb icon

Weighted average cost per Intelligence Index task. Each evaluation’s cost is calculated from input, cache hit, cache write, reasoning, and answer token prices, divided by task count, and weighted by its Intelligence Index weight.

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Tokenverbrauch

Output Tokens per Intelligence Index Task

Weighted average number of output tokens used to run one task in the Artificial Analysis Intelligence Index
Reasoning models are indicated by a lightbulb icon

The number of tokens required per Intelligence Index task. This is calculated by multiplying the output tokens per eval by the relative weights of each benchmark in the Intelligence Index, then dividing by task count (excluding repeats).

Kosten

Cost per Intelligence Index Task

Weighted average cost (USD) per Artificial Analysis Intelligence Index task, segmented by token type. Lower is better
Reasoning models are indicated by a lightbulb icon

Weighted average cost per Intelligence Index task. Each evaluation’s cost is calculated from input, cache hit, cache write, reasoning, and answer token prices, divided by task count, and weighted by its Intelligence Index weight.

Cost to Run Artificial Analysis Intelligence Index

Cost (USD) to run all evaluations in the Artificial Analysis Intelligence Index
Reasoning models are indicated by a lightbulb icon

The cost to run the evaluations in the Artificial Analysis Intelligence Index, calculated using the model's input, cache hit, cache write, reasoning, and answer token prices and the number of tokens used across evaluations (excluding repeats).

Pricing: Cache Hit, Input, and Output

Price (USD per M Tokens)
Reasoning models are indicated by a lightbulb icon

Price per token for cached prompts (previously processed), typically offering a significant discount compared to regular input price, represented as USD per million tokens. The values shown here are the cache hit price; cache write and cache storage are billed separately and vary by provider — see "Cache pricing by provider" for detail.

Kontextfenster

Context Window

Context window: tokens limit · Higher is better
Reasoning models are indicated by a lightbulb icon

Larger context windows are relevant to RAG (Retrieval Augmented Generation) LLM workflows which typically involve reasoning and information retrieval of large amounts of data.

Maximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).

Geschwindigkeit

Gemessen anhand der Ausgabegeschwindigkeit (Tokens pro Sekunde)

Output Speed

Output tokens per second · Higher is better
Reasoning models are indicated by a lightbulb icon

Tokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).

Figures represent performance of the model's first-party API (e.g. OpenAI for o1) or the median across providers where a first-party API is not available (e.g. Meta's Llama models).

Time per Intelligence Index Task

Weighted average decode time (minutes) per task; excludes TTFT and overhead time · Lower is better
Reasoning models are indicated by a lightbulb icon

The weighted average time (seconds) per Artificial Analysis Intelligence Index task. This is calculated by dividing output tokens per task by output speed, weighted by the relative weights of each benchmark in the Intelligence Index.

Latenz

Gemessen anhand der Zeit (in Sekunden) bis zum ersten Token

Latency: Time To First Answer Token

Seconds to first answer token received · Accounts for reasoning model 'thinking' time
Reasoning models are indicated by a lightbulb icon

Time to first answer token received, in seconds, after API request sent. For reasoning models, this includes the 'thinking' time of the model before providing an answer. For models which do not support streaming, this represents time to receive the completion.

Ende-zu-Ende-Antwortzeit

Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed

End-to-End Response Time

Seconds to output 500 tokens, including reasoning model 'thinking' time · Lower is better
Reasoning models are indicated by a lightbulb icon

Seconds to receive a 500 token response. Key components:

  • Input time: Time to receive the first response token
  • Thinking time (only for reasoning models): Time reasoning models spend outputting tokens to reason prior to providing an answer. Amount of tokens based on the average reasoning tokens across a diverse set of 60 prompts (methodology details).
  • Answer time: Time to generate 500 output tokens, based on output speed

Figures represent performance of the model's first-party API (e.g. OpenAI for o1) or the median across providers where a first-party API is not available (e.g. Meta's Llama models).

Modellgröße (nur Modelle mit offenen Gewichten)

Model Size: Total and Active Parameters

Comparison between total model parameters and parameters active during inference
Reasoning models are indicated by a lightbulb icon

The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.

The number of parameters actually executed during each inference forward pass, expressed in billions. For Mixture of Experts (MoE) models, a routing mechanism selects a subset of experts per token, resulting in fewer active than total parameters. Dense models use all parameters, so active equals total.

Häufig gestellte Fragen

Häufige Fragen zu Devstral 2

Devstral 2 wurde am 9. Dezember 2025 veröffentlicht.

Devstral 2 wurde von Mistral entwickelt.

Devstral 2 erzielt im Artificial Analysis Intelligence Index einen Wert von 19 und liegt damit im Vergleich zu anderen Modellen ohne Reasoning mit offenen Gewichten ähnlicher Größe deutlich über dem Durchschnitt (Median: 7).

Devstral 2 erzeugt Ausgaben mit 42.8 Tokens pro Sekunde (auf Grundlage der API von Mistral). Dies ist am unteren Ende im Vergleich zu anderen Modellen ohne Reasoning mit offenen Gewichten ähnlicher Größe (Median: 95.7 Tokens/s).

Devstral 2 hat eine Zeit bis zum ersten Token (TTFT) von 1.47 s (auf Grundlage der API von Mistral). Dies ist besser als der Durchschnitt im Vergleich zu anderen Modellen ohne Reasoning mit offenen Gewichten ähnlicher Größe (Median: 1.77 s).

Bei der Evaluation des Intelligence Index erzeugte Devstral 2 9.2M Ausgabetokens. Dies ist besser als der Durchschnitt im Vergleich zu anderen Modellen ohne Reasoning mit offenen Gewichten ähnlicher Größe (Median: 14M).

Nein, Devstral 2 ist kein Reasoning-Modell. Es antwortet direkt, ohne längeres Chain-of-Thought-Reasoning.

Devstral 2 unterstützt Text als Eingabe.

Devstral 2 unterstützt Text als Ausgabe.

Nein, Devstral 2 unterstützt keine Bildeingaben. Es kann nur Text verarbeiten.

Nein, Devstral 2 ist nicht multimodal. Es unterstützt nur Text als Eingabe.

Devstral 2 hat ein Kontextfenster von 260k Tokens. Es bestimmt, wie viel Text und Gesprächsverlauf das Modell in einer einzelnen Anfrage verarbeiten kann.

Ja, Devstral 2 hat offene Gewichte. Die Modellgewichte sind öffentlich verfügbar und können zum Selbsthosten heruntergeladen werden.

Devstral 2 hat 125 Milliarden Parameter.

Devstral 2 wird unter der Lizenz Modified MIT License veröffentlicht. Für die kommerzielle Nutzung ist eine separate Lizenzvereinbarung erforderlich. Lizenz anzeigen

Devstral 2 erzielt im Artificial Analysis Intelligence Index einen Wert von 19. Dieser zusammengesetzte Benchmark bewertet Modelle in den Bereichen Schlussfolgern, Wissen, Mathematik und Programmierung.

Ja, Devstral 2 ist über 1 Anbieter per API verfügbar. API-Anbieter vergleichen

Devstral 2 ist über 1 API-Anbieter verfügbar. Anbieter vergleichen