Skip to content

Mistral · Model profile

Mistral Medium 3

Intelligence, Performance & Price Analysis

Source: Canonical slug: mistral-medium-3 · Canonical model registry at build time

Routing verdict

Capable everyday model

Good for routine tasks; route complex reasoning and premium workloads to stronger models.

Mistral Medium 3 summary metrics

Intelligence
12 scoreMistral Medium 3 scores 12 (estimated) on the Artificial Analysis Intelligence Index, placing it above average among other non-reasoning models in a similar price tier (median: 11).
Speed
48.2 output tokens/secMistral Medium 3 generates output at 48.2 tokens per second (based on Mistral's API), which is at the lower end compared to other non-reasoning models in a similar price tier (median: 95.5 t/s).
Latency
1.47s TTFTMistral Medium 3 has a time to first token (TTFT) of 1.47s (based on Mistral's API), which is better than average compared to other non-reasoning models in a similar price tier (median: 1.53s).
Input Price
$0.40 / 1M tokensMistral Medium 3 costs $0.40 per 1M input tokens (somewhat higher than average, median: $0.20) and $2.00 per 1M output tokens (at the higher end, median: $0.70), based on Mistral's API.
Output Price
$2.00 / 1M tokensMistral Medium 3 costs $0.40 per 1M input tokens (somewhat higher than average, median: $0.20) and $2.00 per 1M output tokens (at the higher end, median: $0.70), based on Mistral's API.

Assessment

Routing verdict and task fit

Mistral Medium 3 scores 12 (estimated) on the Artificial Analysis Intelligence Index, placing it above average among other non-reasoning models in a similar price tier (median: 11). Mistral Medium 3 generates output at 48.2 tokens per second (based on Mistral's API), which is at the lower end compared to other non-reasoning models in a similar price tier (median: 95.5 t/s). Mistral Medium 3 costs $0.40 per 1M input tokens (somewhat higher than average, median: $0.20) and $2.00 per 1M output tokens (at the higher end, median: $0.70), based on Mistral's API.
Task fit assessment
WorkloadRatingNotes
Complex reasoning & agentic workflowsViableIntelligence score 12 handles routine reasoning but may struggle with open-ended agentic tasks.
High-volume chat & customer-facingViableOutput speed may be a bottleneck for real-time chat at scale.
Latency-sensitive applicationsOptimalTTFT 1.47s — adequate latency for most interactive use cases.
Cost-sensitive pipelinesViableOutput pricing at $2.00 is premium. Route high-volume simple tasks to cheaper alternatives.

Evidence

Evidence charts

14 published comparison charts for this profile. Mistral Medium 3 is marked with a diamond and bold label; each chart has a table view and lists its source and retrieval date.

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. · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3: -31.5, row 11 of 20 in the published order.
  1. Claude Fable 5 (with fallback)40.1
  2. Gemini 3.5 Flash22.7
  3. GPT-5.5 (xhigh)20.1
  4. Grok 4.3 (high)18.3
  5. Kimi K2.66.4
  6. Muse Spark4.1
  7. GLM-5.2 (max)4
  8. MiniMax-M31.4
  9. Nemotron 3 Ultra-0.833
  10. DeepSeek V4 Pro (max)-10
  11. Mistral Medium 3 (this model)-31.5
  12. Mistral Large 3-39.4
  13. DeepSeek V3 (Dec)-41.4
  14. Llama 4 Maverick-41.8
  15. Mistral Medium 3.1 (this model)-45.5
  16. MiMo-V2-Flash-48.5
  17. gpt-oss-120b (high)-50
  18. GPT-4.1 mini-50.1
  19. Mistral Small 3.1-50.3
  20. Mistral Small 3.2-50.4

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

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 · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3.1: 14.7, row 17 of 20 in the published order.
  1. Claude Fable 5 (with fallback)59.9
  2. GPT-5.5 (xhigh)54.8
  3. GLM-5.2 (max)51.1
  4. Gemini 3.5 Flash50.2
  5. MiniMax-M344.4
  6. DeepSeek V4 Pro (max)44.3
  7. Kimi K2.644.2
  8. Muse Spark43.1
  9. Nemotron 3 Ultra37.8
  10. Grok 4.3 (high)37.6
  11. MiMo-V2-Flash24.7
  12. gpt-oss-120b (high)23.8
  13. Qwen3 Coder Next21.1
  14. Mistral Large 315.9
  15. GPT-4.1 mini14.8
  16. Mistral Small 3.114.7
  17. Mistral Medium 3.1 (this model)14.7
  18. Llama 4 Maverick14.3
  19. DeepSeek V3 (Dec)14.2
  20. Ling 2.6 Flash14.1

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

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 · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3.1: 14.7, row 17 of 20 in the published order.
  1. Claude Fable 5 (with fallback)59.9
  2. GPT-5.5 (xhigh)54.8
  3. GLM-5.2 (max)51.1
  4. Gemini 3.5 Flash50.2
  5. MiniMax-M344.4
  6. DeepSeek V4 Pro (max)44.3
  7. Kimi K2.644.2
  8. Muse Spark43.1
  9. Nemotron 3 Ultra37.8
  10. Grok 4.3 (high)37.6
  11. MiMo-V2-Flash24.7
  12. gpt-oss-120b (high)23.8
  13. Qwen3 Coder Next21.1
  14. Mistral Large 315.9
  15. GPT-4.1 mini14.8
  16. Mistral Small 3.114.7
  17. Mistral Medium 3.1 (this model)14.7
  18. Llama 4 Maverick14.3
  19. DeepSeek V3 (Dec)14.2
  20. Ling 2.6 Flash14.1

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Intelligence

Artificial Analysis Intelligence Index · Higher is better · Evaluation results measured independently by Artificial Analysis

12 models in the published order. Mistral Medium 3: 12.5, row 12 of 12 in the published order.
  1. Claude Fable 5 (with fallback)59.9
  2. GPT-5.5 (xhigh)54.8
  3. GLM-5.2 (max)51.1
  4. Gemini 3.5 Flash50.2
  5. MiniMax-M344.4
  6. DeepSeek V4 Pro (max)44.3
  7. Kimi K2.644.2
  8. Muse Spark43.1
  9. Nemotron 3 Ultra37.8
  10. Grok 4.3 (high)37.6
  11. gpt-oss-120b (high)23.8
  12. Mistral Medium 3 (this model)12.5

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Output Speed

Output tokens per second · Higher is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3.1: 89 t/s, row 14 of 20 in the published order.
  1. gpt-oss-120b (high)314 t/s
  2. Nemotron 3 Ultra250 t/s
  3. GLM-5.2 (max)218 t/s
  4. Gemini 3.5 Flash192 t/s
  5. GPT-4.1 nano188 t/s
  6. Ling 2.6 Flash177 t/s
  7. Mistral Small 3.1171 t/s
  8. Grok 4.3 (high)164 t/s
  9. Mistral Small 3.2147 t/s
  10. Llama 4 Maverick118 t/s
  11. Llama 4 Scout109 t/s
  12. MiniMax-M399 t/s
  13. Ministral 3 8B93 t/s
  14. Mistral Medium 3.1 (this model)89 t/s
  15. Qwen3 Coder Next89 t/s
  16. GPT-5.5 (xhigh)88 t/s
  17. GPT-4.1 mini86 t/s
  18. Llama 3.3 70B86 t/s
  19. Ministral 3 14B78 t/s
  20. Kimi K2.676 t/s

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Output speed context
Speed should be read together with latency and cost; fast output does not always mean better user-perceived performance.

Speed

Output tokens per second · Higher is better · Evaluation results measured independently by Artificial Analysis

11 models in the published order. Mistral Medium 3: 48 t/s, row 11 of 11 in the published order.
  1. gpt-oss-120b (high)314 t/s
  2. Nemotron 3 Ultra250 t/s
  3. GLM-5.2 (max)218 t/s
  4. Gemini 3.5 Flash192 t/s
  5. Grok 4.3 (high)164 t/s
  6. MiniMax-M399 t/s
  7. GPT-5.5 (xhigh)88 t/s
  8. Kimi K2.676 t/s
  9. DeepSeek V4 Pro (max)72 t/s
  10. Claude Fable 5 (with fallback)71 t/s
  11. Mistral Medium 3 (this model)48 t/s

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Output speed context
Speed should be read together with latency and cost; fast output does not always mean better user-perceived performance.

End-to-End Response Time

Seconds to output 500 tokens, including reasoning model 'thinking' time · Lower is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3.1: 5.64 s, row 9 of 20 in the published order.
  1. GPT-4.1 nano2.65 s
  2. Mistral Small 3.12.92 s
  3. Mistral Small 3.23.39 s
  4. Ling 2.6 Flash2.82 s
  5. Llama 4 Maverick4.24 s
  6. Llama 4 Scout4.61 s
  7. Ministral 3 8B5.35 s
  8. GPT-4.1 mini5.84 s
  9. Mistral Medium 3.1 (this model)5.64 s
  10. Ministral 3 14B6.4 s
  11. Llama 3.3 70B5.85 s
  12. Qwen3 Coder Next5.65 s
  13. gpt-oss-120b (high)1.59 s
  14. Mistral Large 39.43 s
  15. Mistral Medium 3 (this model)10.38 s
  16. Nemotron 3 Ultra2 s
  17. GLM-5.2 (max)2.3 s
  18. Grok 4.3 (high)3.05 s
  19. MiniMax-M35.05 s
  20. Gemini 3.5 Flash2.6 s

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

Latency: Time To First Answer Token

Seconds to first answer token received · Accounts for reasoning model 'thinking' time · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3: 0 s, row 11 of 20 in the published order.
  1. GPT-4.1 nano0 s
  2. Mistral Small 3.20 s
  3. Mistral Small 3.10 s
  4. Ministral 3 8B0 s
  5. Llama 4 Scout0 s
  6. Ministral 3 14B0 s
  7. Llama 4 Maverick0 s
  8. GPT-4.1 mini0 s
  9. Mistral Large 30 s
  10. Ling 2.6 Flash0 s
  11. Mistral Medium 3 (this model)0 s
  12. Mistral Medium 3.1 (this model)0 s
  13. Llama 3.3 70B0 s
  14. Qwen3 Coder Next0 s
  15. gpt-oss-120b (high)6.37 s
  16. Nemotron 3 Ultra9.09 s
  17. GLM-5.2 (max)9.19 s
  18. Grok 4.3 (high)0 s
  19. MiniMax-M320.19 s
  20. Gemini 3.5 Flash0 s

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Latency considerations
Latency includes thinking / reasoning time when applicable. Lower is better for interactive use cases.

Context Window

Context window: tokens limit · Higher is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3 is not marked in this published set of 20 models.
  1. Llama 4 Scout10M
  2. Gemini 3.5 Flash1M
  3. Claude Fable 5 (with fallback)1M
  4. GLM-5.2 (max)1M
  5. DeepSeek V4 Pro (max)1M
  6. Grok 4.3 (high)1M
  7. MiniMax-M31M
  8. GPT-4.1 mini1M
  9. Llama 4 Maverick1M
  10. GPT-4.1 nano1M
  11. GPT-5.5 (xhigh)922k
  12. Nemotron 3 Ultra262.1k
  13. Muse Spark262.1k
  14. Ling 2.6 Flash262.1k
  15. Kimi K2.6256k
  16. MiMo-V2-Flash256k
  17. Qwen3 Coder Next256k
  18. Mistral Large 3256k
  19. Ministral 3 14B256k
  20. Ministral 3 8B256k

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Context window sizing
Larger context windows enable complex multi-turn tasks but may increase latency and cost for long inputs.

Cost per Intelligence Index Task

Weighted average cost (USD) per Artificial Analysis Intelligence Index task, segmented by token type. Lower is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3.1: 0.0101, row 13 of 20 in the published order.
  1. Llama 4 Scout0.000824
  2. DeepSeek V3 (Dec)0.00202
  3. Llama 4 Maverick0.00178
  4. Mistral Small 3.10.00143
  5. DeepSeek V4 Pro (max)0.0063
  6. GPT-4.1 mini0.00473
  7. Mistral Large 30.00455
  8. gpt-oss-120b (high)0.00454
  9. Llama 3.3 70B0.0012
  10. Mistral Small 3.20.00242
  11. MiniMax-M30.0139
  12. Grok 4.3 (high)0.0101
  13. Mistral Medium 3.1 (this model)0.0101
  14. Ministral 3 14B0.00235
  15. Ministral 3 8B0.00235
  16. Nemotron 3 Ultra0.0209
  17. Qwen3 Coder Next0.0184
  18. Kimi K2.60.0216
  19. MiMo-V2-Flash0.00812
  20. GLM-5.2 (max)0.0268

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Cost pressure
Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.

Cost per Task

Weighted average cost (USD) per Intelligence Index task · Lower is better · Evaluation results measured independently by Artificial Analysis

10 models in the published order. Mistral Medium 3 is not marked in this published set of 10 models.
  1. DeepSeek V4 Pro (max)$0.045
  2. gpt-oss-120b (high)$0.061
  3. MiniMax-M3$0.125
  4. Grok 4.3 (high)$0.139
  5. Nemotron 3 Ultra$0.244
  6. Kimi K2.6$0.346
  7. GLM-5.2 (max)$0.370
  8. Gemini 3.5 Flash$0.586
  9. GPT-5.5 (xhigh)$0.863
  10. Claude Fable 5 (with fallback)$2.75

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Cost pressure
Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.

Cost to Run Artificial Analysis Intelligence Index

Cost (USD) to run all evaluations in the Artificial Analysis Intelligence Index · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3.1: 17.3, row 12 of 20 in the published order.
  1. Llama 4 Scout2.7
  2. DeepSeek V3 (Dec)2.9
  3. Mistral Small 3.11.5
  4. Llama 4 Maverick4
  5. GPT-4.1 mini8.3
  6. Mistral Large 38.4
  7. Llama 3.3 70B2.7
  8. gpt-oss-120b (high)3.5
  9. Ministral 3 8B2.8
  10. Ministral 3 14B3
  11. Mistral Small 3.22.5
  12. Mistral Medium 3.1 (this model)17.3
  13. DeepSeek V4 Pro (max)7.2
  14. MiniMax-M314.5
  15. Qwen3 Coder Next38.5
  16. Grok 4.3 (high)12.5
  17. MiMo-V2-Flash11.4
  18. Nemotron 3 Ultra33.5
  19. GLM-5.2 (max)32.9
  20. Kimi K2.625

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Cost pressure
Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.

Pricing: Cache Hit, Input, and Output

Price (USD per M Tokens) · Evaluation results measured independently by Artificial Analysis

20 models; segments: Input, Output, Cache Hit. Mistral Medium 3: $2.40, row 18 of 20 in the published order.
  • Input
  • Output
  • Cache Hit
  1. Ministral 3 8B$0.300
  2. MiMo-V2-Flash$0.400
  3. Mistral Small 3.1$0.400
  4. Ling 2.6 Flash$0.420
  5. Ministral 3 14B$0.400
  6. Mistral Small 3.2$0.400
  7. GPT-4.1 nano$0.525
  8. gpt-oss-120b (high)$0.900
  9. Llama 4 Scout$0.830
  10. Llama 4 Maverick$1.47
  11. DeepSeek V3 (Dec)$1.29
  12. Llama 3.3 70B$1.87
  13. DeepSeek V4 Pro (max)$1.31
  14. MiniMax-M3$1.56
  15. Qwen3 Coder Next$1.90
  16. Mistral Large 3$2.00
  17. GPT-4.1 mini$2.10
  18. Mistral Medium 3 (this model)$2.40
  19. Mistral Medium 3.1 (this model)$2.40
  20. Nemotron 3 Ultra$3.60

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

Time per Intelligence Index Task

Weighted average decode time (minutes) per task; excludes TTFT and overhead time · Lower is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Mistral Medium 3.1: 0.935, row 9 of 20 in the published order.
  1. Llama 4 Scout0.198
  2. Llama 4 Maverick0.3
  3. Llama 3.3 70B0.321
  4. Mistral Small 3.10.447
  5. GPT-4.1 nano0.559
  6. GPT-4.1 mini0.575
  7. Mistral Small 3.20.87
  8. Mistral Large 30.924
  9. Mistral Medium 3.1 (this model)0.935
  10. Ling 2.6 Flash1.1
  11. Nemotron 3 Ultra1.4
  12. Grok 4.3 (high)1.6
  13. gpt-oss-120b (high)1.9
  14. Gemini 3.5 Flash2.2
  15. Ministral 3 14B2.2
  16. Ministral 3 8B2.5
  17. Qwen3 Coder Next2.6
  18. GLM-5.2 (max)2.8
  19. GPT-5.5 (xhigh)3
  20. MiniMax-M33.9

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Specifications

Architecture and limits

Mistral Medium 3 technical specifications
SpecificationValueSource
Model typeProprietaryfaq
ReasoningNofaq
Input modalitiesMistral Medium 3 supports text and image input.faq
Output modalitiesMistral Medium 3 supports text output.faq
Context window130k tokensfaq
Open weights / sourceNo, Mistral Medium 3 is proprietary. The model weights are not publicly available.faq
ParametersMistral Medium 3 is a proprietary model and Mistral has not disclosed the model size or parameter count.faq
API availabilityYes, Mistral Medium 3 is available via API through 2 providers.faq

Method

Methodology & Provenance

This page is rendered from the normalized profile and page JSON for Mistral Medium 3.

Benchmark values are preserved as normalized; only layout, disclosure ordering, and typography are adjusted for readability.

Questions

Model FAQ and technical disclosures

Mistral Medium 3 was released on May 7, 2025.

Mistral Medium 3 was created by Mistral.

Mistral Medium 3 scores 12 (estimated) on the Artificial Analysis Intelligence Index, placing it above average among other non-reasoning models in a similar price tier (median: 11).

Mistral Medium 3 generates output at 48.2 tokens per second (based on Mistral's API), which is at the lower end compared to other non-reasoning models in a similar price tier (median: 95.5 t/s).

Mistral Medium 3 has a time to first token (TTFT) of 1.47s (based on Mistral's API), which is better than average compared to other non-reasoning models in a similar price tier (median: 1.53s).

Mistral Medium 3 costs $0.40 per 1M input tokens (somewhat higher than average, median: $0.20) and $2.00 per 1M output tokens (at the higher end, median: $0.70), based on Mistral's API.

Mistral Medium 3 costs $0.40 per 1M input tokens and $2.00 per 1M output tokens (based on Mistral's API). For a blended rate (7:2:1 cache hit/input/output ratio), this is $0.56 per 1M tokens. Pricing may vary by provider.

No, Mistral Medium 3 is not a reasoning model. It provides direct responses without extended chain-of-thought reasoning.

Mistral Medium 3 supports text and image input.

Mistral Medium 3 supports text output.

Yes, Mistral Medium 3 supports image input and can analyze, describe, and answer questions about images.

Yes, Mistral Medium 3 is multimodal. It can process text and image input and generate text output.

Mistral Medium 3 has a context window of 130k tokens. This determines how much text and conversation history the model can process in a single request.

No, Mistral Medium 3 is proprietary. The model weights are not publicly available.

Mistral Medium 3 is a proprietary model and Mistral has not disclosed the model size or parameter count.

Mistral Medium 3 achieves a score of 12 on the Artificial Analysis Intelligence Index. This composite benchmark evaluates models across reasoning, knowledge, mathematics, and coding.

Yes, Mistral Medium 3 is available via API through 2 providers.

Mistral Medium 3 is available through 2 API providers.

Models

  • 01MistralOther Mistral models in the catalog.
  • 02LeaderboardsCertified metric ranks, when published.