AI 模型情报

厂商 · 2026-06-29

Mistral AI

European LLM lab — Mistral, Mixtral, Codestral, Devstral.

48 个标准模型48 条记录(含衍生 / 微调)官方站点 →
模型输入 / 1M输出 / 1M上下文服务商标签
Voxtral Small 24B 2507$0.002$0.00232K4tools · json · open-weights
Mistral Nemo$0.020$0.040128K6tools · open-weights
Ministral 3B (latest)$0.040$0.040128K1tools · open-weights
Mistral Embed$0.100Unknown8K2
Ministral 3 3B 2512$0.100$0.100131K3tools · json · vision · open-weights
Ministral 8B (latest)$0.100$0.100128K1tools · open-weights
Mistral Small 3.2 24B$0.060$0.180128K3tools · json · vision · open-weights
Mistral Nemo$0.150$0.150128K9tools · open-weights
Ministral 3 8B 2512$0.150$0.150262K3tools · json · vision · open-weights
Pixtral 12B$0.150$0.150128K2tools · vision · open-weights
Open Mistral Nemo$0.150$0.150128K1tools · open-weights
Mistral Small 3.1$0.100$0.300128K4tools · vision
Devstral Small$0.100$0.300128K3tools · open-weights
Devstral Small 2505$0.100$0.300128K3tools · open-weights
Ministral 3 14B 2512$0.200$0.200262K3tools · json · vision · open-weights
Mistral Small 3.2$0.100$0.300128K2tools · vision · open-weights
Mistral Small 3.2 24B Instruct 2506$0.100$0.31032K5tools · json · reasoning · open-weights
Mistral 7B$0.250$0.2508K1tools · open-weights
Mistral Small 4$0.150$0.600256K4tools · reasoning · vision · open-weights
Mistral Small (latest)$0.150$0.600256K2tools · reasoning · vision · open-weights
Saba$0.199$0.59533K3tools · json
Mistral Small 3.1 24B$0.351$0.555128K4vision · open-weights
Codestral 2508$0.300$0.900256K5tools · json
Codestral 25.01$0.300$0.900256K3tools
Codestral (latest)$0.300$0.900256K2tools · open-weights
Mixtral 8x7B$0.700$0.70032K1tools · open-weights
Mistral Large 3$0.500$1.50262K8tools · vision · open-weights
Mistral Large (latest)$0.500$1.50262K3tools · vision · open-weights
Magistral Small$0.500$1.50128K2tools · reasoning · open-weights
Devstral 2$0.400$2.00262K7tools · open-weights
Mistral Medium 3$0.400$2.00131K6tools · vision
Mistral Medium 3.1$0.400$2.00131K3tools · json · vision
Mistral Medium 3$0.400$2.00131K3tools · json · vision
Mistral Medium (latest)$0.400$2.00262K2tools · vision
Devstral 2 (latest)$0.400$2.00262K2tools · open-weights
Devstral Medium$0.400$2.00128K2tools · open-weights
Devstral 2$0.400$2.00262K1tools · open-weights
Mistral Medium 3.1$0.400$2.00262K1tools · vision
Magistral Medium (latest)$2.00$5.00128K2tools · reasoning · open-weights
Mistral Medium 3.5 128B$1.65$5.50262K3tools · json · reasoning · vision · open-weights
Mistral Large 2.1$2.00$6.00131K7tools · open-weights
Mixtral 8x22B Instruct$2.00$6.0066K3tools · json · open-weights
Mistral Large$2.00$6.00128K3tools · json
Pixtral Large (latest)$2.00$6.00128K2tools · vision · open-weights
Mixtral 8x22B$2.00$6.0064K1tools · open-weights
Mistral Medium 3.5$1.50$7.50262K4tools · json · reasoning · vision
Mistral Medium 3.5$1.50$7.50262K1tools · json · reasoning · vision · open-weights
Devstral Small 2未公开未公开256K1tools · vision · open-weights

Frequently asked questions

How many AI models does Mistral offer?

We track 48 canonical Mistral models. The list is recomputed daily.

Which Mistral model is the cheapest?

Voxtral Small 24B 2507 is currently the lowest-priced Mistral model, at $0.002 per 1M input tokens and $0.002 per 1M output tokens. For the full apples-to-apples list, see /pricing/cheapest-llm-api.

Which Mistral model has the largest context window?

Ministral 3 8B 2512 leads at 262K tokens. This is the total of prompt + completion.

Which Mistral models support tool calling?

Multiple Mistral models support tool calling, with Voxtral Small 24B 2507 being a popular pick. The capability column in the table above marks every model with Mistral tool-calling support.

Which Mistral models accept image input?

Ministral 3 3B 2512 accepts image input. Other vision-capable Mistral models are tagged 'vision' in the table above. See /capabilities/vision for a cross-vendor comparison.

What are the best alternatives to Mistral?

Depends on the use case. For raw cost savings, look at /pricing/cheapest-llm-api. For agent-oriented workloads, /best/best-ai-model-for-agents. For long-document workflows, /best/best-long-context-llm.

How fresh is this Mistral pricing data?

Daily. Our pipeline syncs every morning and rebuilds these pages on data change, so list-price moves and new model releases land within roughly 24 hours.

最近更新:

Prices in USD per 1M tokens. Unknown means the provider does not publish per-token pricing.

Pricing and capabilities are refreshed daily and reconciled against each provider's official documentation. Always verify critical production decisions with the provider directly.