Mistral Small 4 Launches as Europe's Versatile AI Powerhouse
In the competitive landscape of open-source large language models, European standout Mistral AI has once again showcased its impressive pace of innovation.
On March 16, local time, Mistral AI officially launched Mistral Small4. This marks the lab's first truly "generalist" large model, successfully integrating flagship-level reasoning, multimodal understanding, and robust coding capabilities into a single architecture for the first time. For developers, this eliminates the need to choose between specialized vertical models, as the new Small4 delivers comprehensive performance.

Mistral Small4 utilizes an advanced Mixture of Experts (MoE) architecture:
Core Parameters: The model boasts a total of 119B parameters, with only 6B active during inference, significantly boosting operational efficiency without compromising capability.
Extended Context: It features an extended context window of 256k tokens, enabling it to process entire technical documents or extensive code repositories with ease.
Flexible Modes: It supports both fast-response and deep-reasoning modes and is openly released under the permissive Apache 2.0 license.
In terms of performance, Mistral Small4 represents a substantial leap over its predecessor. Official data indicates a 40% reduction in end-to-end completion time in latency-optimized mode. In throughput-optimized mode, it can handle three times as many requests per second as Small3. In cross-evaluations against other leading models, its performance on three key benchmarks is competitive with OpenAI's GPT-OSS120B.
Deployment Requirements and Hardware Recommendations:
To fully leverage this model's potential, Mistral AI provides clear hardware guidance. The minimum recommended setup is 4× HGX H100 or 1× DGX B200. For optimal performance, the official recommendation is a configuration using 4× HGX H200 or 2× DGX B200.
With the release of Mistral Small4, Mistral AI
Related article
Stripe to Acquire OpenRouter as AI Model Routing Gains Momentum
Stripe has agreed to acquire OpenRouter, an AI model-routing platform that gives developers access to hundreds of models through a single interface. The deal adds model selection and routing to Stripe’s existing work around AI usage and token-based b
Former OpenAI Researcher Barrett Zoph Returns to Google to Support Gemini Development
Barret Zoph, who previously abruptly left and joined OpenAI due to a disagreement with Thinking Machines Lab CEO Miriam Mularati, has once again changed employers. Alphabet's Google announced that Zoph will return as Vice President of Research, bring
How Recursive Intelligence Raised $335M at $4B Valuation in 4 Months
The co-founders of startup Ricursive Intelligence seemed destined to be co-founders.Anna Goldie, CEO, and Azalia Mirhoseini, CTO, are so well-known in the AI community that they were among those AI engineers who “got those weird emails from Zuckerber
Related Special Topic Recommendations
Comments (1)
0/500
Mistral Small klingt echt vielversprechend! Als Europäer freut es mich, dass hier nicht nur die großen US-Firmen die Szene dominieren. Hoffentlich bleibt das Modell auch wirklich 'open source' und wird nicht irgendwann eingeschränkt. Die Geschwindigkeit, mit der Mistral neue Modelle raushaut, ist schon beeindruckend – fast ein bisschen beängstigend. 🧐 Kann es in der Praxis mit den etablierten Größen mithalten?
In the competitive landscape of open-source large language models, European standout Mistral AI has once again showcased its impressive pace of innovation.
On March 16, local time, Mistral AI officially launched Mistral Small4. This marks the lab's first truly "generalist" large model, successfully integrating flagship-level reasoning, multimodal understanding, and robust coding capabilities into a single architecture for the first time. For developers, this eliminates the need to choose between specialized vertical models, as the new Small4 delivers comprehensive performance.

Mistral Small4 utilizes an advanced Mixture of Experts (MoE) architecture:
Core Parameters: The model boasts a total of 119B parameters, with only 6B active during inference, significantly boosting operational efficiency without compromising capability.
Extended Context: It features an extended context window of 256k tokens, enabling it to process entire technical documents or extensive code repositories with ease.
Flexible Modes: It supports both fast-response and deep-reasoning modes and is openly released under the permissive Apache 2.0 license.
In terms of performance, Mistral Small4 represents a substantial leap over its predecessor. Official data indicates a 40% reduction in end-to-end completion time in latency-optimized mode. In throughput-optimized mode, it can handle three times as many requests per second as Small3. In cross-evaluations against other leading models, its performance on three key benchmarks is competitive with OpenAI's GPT-OSS120B.
Deployment Requirements and Hardware Recommendations:
To fully leverage this model's potential, Mistral AI provides clear hardware guidance. The minimum recommended setup is 4× HGX H100 or 1× DGX B200. For optimal performance, the official recommendation is a configuration using 4× HGX H200 or 2× DGX B200.
With the release of Mistral Small4, Mistral AI
Stripe to Acquire OpenRouter as AI Model Routing Gains Momentum
Stripe has agreed to acquire OpenRouter, an AI model-routing platform that gives developers access to hundreds of models through a single interface. The deal adds model selection and routing to Stripe’s existing work around AI usage and token-based b
Former OpenAI Researcher Barrett Zoph Returns to Google to Support Gemini Development
Barret Zoph, who previously abruptly left and joined OpenAI due to a disagreement with Thinking Machines Lab CEO Miriam Mularati, has once again changed employers. Alphabet's Google announced that Zoph will return as Vice President of Research, bring
How Recursive Intelligence Raised $335M at $4B Valuation in 4 Months
The co-founders of startup Ricursive Intelligence seemed destined to be co-founders.Anna Goldie, CEO, and Azalia Mirhoseini, CTO, are so well-known in the AI community that they were among those AI engineers who “got those weird emails from Zuckerber
Mistral Small klingt echt vielversprechend! Als Europäer freut es mich, dass hier nicht nur die großen US-Firmen die Szene dominieren. Hoffentlich bleibt das Modell auch wirklich 'open source' und wird nicht irgendwann eingeschränkt. Die Geschwindigkeit, mit der Mistral neue Modelle raushaut, ist schon beeindruckend – fast ein bisschen beängstigend. 🧐 Kann es in der Praxis mit den etablierten Größen mithalten?





Home






