2026-06-10 20:57 UTC · mistral/pricing · live page
Summary by fallback. It may be wrong; the diff below is what actually changed. Raw: f32b7be557-mistral-pricing.diff · commit f32b7be557
diff --git a/snapshots/mistral/pricing.md b/snapshots/mistral/pricing.mdindex a4cadd7..764888e 100644--- a/snapshots/mistral/pricing.md+++ b/snapshots/mistral/pricing.md@@ -91,5 +91,5 @@ Start now | | Free Start now | Pro Start now | Team Start now | Enterprise Contact us | |---|---|---|---|---|-| Interfaces | | | | |+| Interfaces. | | | | | | Vibe on web Vibe in your browser: run long-horizon work and code tasks from a chat interface. | | | | | | Vibe CLI Terminal-native coding agent powered by Devstral for autonomous software engineering tasks. | | | | |@@ -156,19 +156,19 @@ Yes, the Mistral Enterprise plan includes custom SLAs, dedicated support, and pr Our most powerful models and APIs. -Open- Mistral Medium 3.5 +Open+ Our flagship model that merges instruction-following, reasoning, and coding into a single 128B dense model. Released as open weights, under a modified MIT license. Text-to-text -Agentic--Multimodal+Reasoning Coding -Reasoning+Agentic++Multimodal Input (/M tokens)@@ -180,10 +180,10 @@ Output (/M tokens) $7.5 -Learn more--Open+mistral-medium-latest Voxtral TTS +Open+ State-of-the-art text-to-speech generation and voice cloning. @@ -194,8 +194,14 @@ Audio generation $0.016 per 1k characters -Premier+Available on++`/v1/audio/speech`++voxtral-mini-tts-latest OCR 3 +Premier+ The world's best document extraction and understanding model. @@ -212,5 +218,5 @@ $2 / 1000 pages Annotations -$3 / 1,000 pages+$3 / 1000 pages mistral-ocr-latest@@ -228,15 +234,15 @@ Cutting-edge AI models across text, reasoning, vision, and more—ready to integ Capabilities -33 results+32 results - Text-to-text-- Multimodal-- Agentic-- Coding - Reasoning-- OCR+- Coding+- Agentic+- Multimodal - Lightweight-- Transcription+- OCR - Voice+- Transcription - Classifier APIs - Embedding@@ -249,29 +255,35 @@ Licenses - Labs -Open+Mistral Medium 3.5 -Mistral Large 3+Open -Open-weight, general-purpose, flagship multimodal and multilingual model.+Our flagship model that merges instruction-following, reasoning, and coding into a single 128B dense model. Released as open weights, under a modified MIT license. Text-to-text +Reasoning++Coding++Agentic+ Multimodal Input (/M tokens) -$0.5+$1.5 Output (/M tokens) -$1.5+$7.5 -mistral-large-latest+mistral-medium-latest -Open+Mistral Small 4 -Mistral Medium 3.5+Open -Our flagship model that merges instruction-following, reasoning, and coding into a single 128B dense model. Released as open weights, under a modified MIT license.+SOTA. Multimodal. Multilingual. Apache 2.0. Text-to-text@@ -281,22 +293,20 @@ Agentic Multimodal -Coding--Reasoning+Lightweight Input (/M tokens) -$1.5+$0.1 Output (/M tokens) -$7.5--Learn more+$0.3 -Premier+mistral-small-latest OCR 3 +Premier+ The world's best document extraction and understanding model. @@ -313,116 +323,144 @@ $2 / 1000 pages Annotations -$3 / 1,000 pages+$3 / 1000 pages mistral-ocr-latest -Open--Devstral 2--Open-weights agentic coding model for autonomous software engineering.+Mistral Large 3 -Coding+Open -Agentic+Open-weight, general-purpose, flagship multimodal and multilingual model. Text-to-text +Multimodal+ Input (/M tokens) -$0.4+$0.5 Output (/M tokens) -$2+$1.5 -devstral-medium-latest+mistral-large-latest -Labs+Voxtral TTS -Devstral Small 2+Open -The best lightweight, open model for coding agents.+State-of-the-art text-to-speech generation and voice cloning. -Coding+Voice -Agentic+Audio generation -Text-to-text+$0.016 per 1k characters++Available on++`/v1/audio/speech`++voxtral-mini-tts-latest++Voxtral Mini Transcribe 2++Premier++State-of-the-art transcription model++Transcription Lightweight -Multimodal+Audio Input/min -Input (/M tokens)+$0.003 -$0.1+Available on -Output (/M tokens)+`/v1/audio/transcriptions` -$0.3+voxtral-mini-latest -devstral-small-latest+Voxtral Mini Transcribe Realtime -Premier+Open -Codestral+Open model fine-tuned and optimized for transcription. -Low-latency coding model optimized for high-frequency completion, fill-in-the-middle, and code generation tasks.+Audio Input/min -Coding+$0.006 -Text-to-text+Available on -Input (/M tokens)+`/v1/audio/transcriptions` -$0.3+voxtral-mini-transcribe-realtime-2602 -Output (/M tokens)+Voxtral Small -$0.9+Open -Training Cost (/M tokens)+State-of-the-art performance on speech and audio understanding. -$3+Transcription -Storage Cost+Text-to-text -Input (/M tokens)+Input (per min / per M tok) -$0.2+$0.004 (audio) / $0.1 (text) Output (/M tokens) -$0.6+$0.4 -codestral-latest+Available on -Labs+`/v1/chat/completions` -Leanstral+voxtral-small-latest -First open-source code agent for Lean 4.+Devstral 2++Open++Open-weights agentic coding model for autonomous software engineering. Coding -API endpoint+Agentic -labs-leanstral-2603+Text-to-text -Open+Input (/M tokens) -Mistral Small 4+$0.4 -SOTA. Multimodal. Multilingual. Apache 2.0.+Output (/M tokens) -Text-to-text+$2++devstral-medium-latest++Devstral Small 2++Labs++The best lightweight, open model for coding agents.++Coding Agentic -Multimodal+Text-to-text Lightweight +Multimodal+ Input (/M tokens) @@ -433,32 +471,44 @@ Output (/M tokens) $0.3 -mistral-small-latest--Labs+devstral-small-latest -Mistral Small Creative+Codestral -A fine-tuned small model for creative writing, roleplay, and chat—trained on curated data.+Premier -Text-to-text+Low-latency coding model optimized for high-frequency completion, fill-in-the-middle, and code generation tasks. -Multimodal+Coding -Lightweight+Text-to-text Input (/M tokens) -$0.1+$0.3 Output (/M tokens) -$0.3+$0.9 -labs-mistral-small-creative+codestral-latest -Premier+Leanstral++Labs++First open-source code agent for Lean 4.++Coding++API endpoint++Free++labs-leanstral-2603 Magistral Medium +Premier+ Thinking model excelling in domain-specific, transparent, and multilingual reasoning. @@ -479,8 +529,8 @@ $5 magistral-medium-latest -Premier- Magistral Small +Premier+ Thinking model excelling in domain-specific, transparent, and multilingual reasoning. @@ -503,8 +553,8 @@ $1.5 magistral-small-latest -Open- Ministral 3 - 3B +Open+ Best-in-class frontier AI to the edge. @@ -525,8 +575,8 @@ $0.1 ministral-3b-latest -Open- Ministral 3 - 8B +Open+ Best-in-class frontier AI to the edge. @@ -547,8 +597,8 @@ $0.15 ministral-8b-latest -Open- Ministral 3 - 14B +Open+ Best-in-class frontier AI to the edge. @@ -569,78 +619,4 @@ $0.2 ministral-14b-latest -Open--Voxtral Small--State-of-the-art performance on speech and audio understanding.--Transcription--Text-to-text--Input (per min / per M tok)--$0.004 (audio) / $0.1 (text)--Output (/M tokens)--$0.4--Available on--voxtral-small-latest--Open--Voxtral Mini--Low-latency speech recognition for edge and devices.--Voice--Text-to-text--Lightweight--Input (per min / per M tok)--$0.001 (audio) / $0.04 (text)--Output (/M tokens)--$0.04--Available on--voxtral-mini-latest--Premier--Voxtral Mini Transcribe--State-of-the-art transcription model.--Voice--Text-to-text--Price--$0.002 per min--voxtral-mini-latest--Open--Voxtral TTS--State-of-the-art text-to-speech generation and voice cloning.--Voice--Audio generation--$0.016 per 1k characters- Classifier API model 3B @@ -703,8 +679,8 @@ $0.1 mistral-moderation-2603 -Premier- Codestral Embed +Premier+ State-of-the-art embeddings for code and natural language queries. @@ -741,5 +717,5 @@ Tools Price -Agent API+Model cost per M token + tool call Libraries@@ -771,6 +747,4 @@ Price (per 1K calls) $30 -Code execution- Web search @@ -783,6 +757,4 @@ Price (per 1K calls) $30 -Web search- Images @@ -795,6 +767,4 @@ Price (per 1K images) $100 -Images- Premium news @@ -807,6 +777,4 @@ Price (per 1K calls) $50 -Premium news- Data capture @@ -819,10 +787,8 @@ Price (per M tokens) $0.04 -Data capture+Mistral NeMo Open -Mistral NeMo- State-of-the-art Mistral model trained specifically for code tasks. @@ -839,24 +805,10 @@ Output (/M tokens) $0.15 -Training Cost (/M tokens)--$1--Storage Cost--Input (/M tokens)--$0.15--Output (/M tokens)--$0.15- open-mistral-nemo -Open- Mixtral 8x7B +Open+ A 7B sparse Mixture-of-Experts (SMoE). Uses 12.9B active parameters out of 45B total. @@ -875,8 +827,8 @@ $0.7 open-mixtral-8x7b -Open- Mixtral 8x22B +Open+ Mixtral 8x22B is currently the most performant open model. A performant, 22B sparse Mixture-of-Experts (SMoE). Uses only 39B active parameters out of 141B. @@ -899,6 +851,4 @@ Our sales experts are here to provide all the details you need. Fill out the for Thank you for your submission. -France- Build your own AI future.