mistral-large-latest

Mistral Large is Mistral AI's flagship model, available in multiple versions with "latest" pointing to the most recent release. Mistral AI has built a reputation for strong performance per parameter and competitive pricing. Mistral Large supports a 128K-token context window, function calling, JSON mode, and streaming. It is available via Mistral's API (La Plateforme) and through major cloud providers. The model is particularly well-regarded for its strong multilingual performance across European languages.

Mistral Large is priced at $2.00 per million input tokens and $6.00 per million output tokens — competitively priced relative to other advanced-tier models. At 1,000 input + 500 output tokens per call, it costs approximately $0.005 per 1,000 API calls. The 3:1 output-to-input ratio is lower than most competitors, making it particularly cost-effective for output-heavy workloads. For European teams with multilingual requirements, Mistral Large's combination of competitive pricing and strong multilingual performance is a compelling value proposition.

Mistral Large's 128K-token context window handles most production use cases. Median latency is approximately 300–500ms TTFT. Throughput is approximately 80–100 tokens per second. Maximum output is 4,096 tokens — lower than many competing models, which can be a constraint for long-form generation tasks. For applications requiring long outputs, this limitation may necessitate multi-call strategies or switching to a model with higher output limits. The model's European data residency options (via Mistral's infrastructure) are relevant for GDPR-sensitive workloads.

Mistral Large is best suited to general-purpose language tasks with multilingual requirements, enterprise applications requiring European data residency, coding assistance, document analysis, and structured output generation. Its function calling implementation is reliable and well-documented. JSON mode enables structured output generation without prompt engineering. For European enterprises with GDPR compliance requirements, Mistral's French infrastructure and data residency options are a significant advantage over US-based providers. The model handles French, Spanish, German, Italian, and Portuguese particularly well — better than most US-based models of comparable size.

Mistral Large sits at the top of Mistral's proprietary lineup, above Mistral Small and Mistral Nemo. Its main competitors are advanced-tier proprietary models from OpenAI and Anthropic (more expensive, larger ecosystems), and open-weights models of similar capability that can be self-hosted. For teams with European data residency requirements or strong multilingual needs, Mistral Large is often the preferred choice. For teams without these specific requirements, other providers offer broader ecosystem support. Its intelligence score of approximately 80/100 reflects solid advanced-tier performance.