Mistral Large
Feb 2024
Mistral LargeCompare inon
Mistral Large measured on 5 of 104 Chat & Knowledge benchmarks5
BenchmarkByScoreRank
Overall1
Knowledge2
Science1
Legacy · saturated or no longer reported1
No benchmark matches
Rankingin non-thinking mode?
Full ranking →···
327Qwen2.5 Coder 32B Instruct3.5328Mistral Large3.5329Granite 4.1 3B3.4Strong at
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperMiMo v2.6 ProthinkingOpen×14.1-82%$0.13Pass back reasoning_content+3
BetterGPT-6 Lunalow×5.8-93%$0.005Function calling only with none · Chat Completions+3
CheaperMuse Spark 1.3max×13.9-33%$1.60Only auto tool_choice is supported+5
CheaperMiniMax M3.1 Flashthinking×13.7-91%$0.22Complete model response must be appended
CheaperGemini 3.8 Flashhigh×13.7-50%*$1.24Sampling parameters are ignored+1ModelScoreCostPer taskWhat changes
MiMo v2.6 ProthinkingOpen×14.1-82%$0.13Pass back reasoning_content+3Muse Spark 1.3max×13.9-33%$1.60Only auto tool_choice is supported+5MiniMax M3.1 Flashthinking×13.7-91%$0.22Complete model response must be appendedGemini 3.8 Flashhigh×13.7-50%*$1.24Sampling parameters are ignored+1
Step 5 PreviewthinkingOpen×13.3-52%$1.03web_search not available · Chat Completions+2ModelScoreCostPer taskWhat changes
GPT-6 Lunalow×5.8-93%$0.005Function calling only with none · Chat Completions+3
Granite 4.2 3BthinkingOpen+89%-98%$0.006GPT-5.6 Lunalow×5.7-85%$0.01GPT OSS 20BhighOpen×3.1-99%$0.011Mistral Small 4thinkingOpen×2.8-91%$0.015ModelScoreCostPer taskWhat changes
MiMo v2.6 ProthinkingOpen×14.1-82%$0.13Pass back reasoning_content+3
Kimi K3maxOpen×13.4+100%$2.00Sampling parameters are fixed+3Step 5 PreviewthinkingOpen×13.3-52%$1.03web_search not available · Chat Completions+2Muse Spark OpenxhighOpen×12.5-33%$0.84Reasoning off leaves output unchanged · Self-hosted / llama.cpp+1
DeepSeek V4.1 Flash 1017maxOpen×12.4-83%$0.26Pass back reasoning_content · Chat Completions+4ModelScoreCostPer taskWhat changes
Mistral Large 4thinking×10—$1.13Infinite whitespace stream+4Mistral Medium 3.5highOpen×3.9—$0.50Mistral Small 4thinkingOpen×2.8—$0.015Magistral Medium 1thinking×2.8——Context: 128K → 40KMagistral Small 1thinkingOpen×2.2—$0.70 / $1.16Ministral 3 3B 2512Open+54%—$0.008Ministral 3 14B 2512non-thinkingOpen+32%—$0.019Ministral 3 8B 2512Opensame—$0.011All Mistral models →History
4 versionsBenchmarks
in Chat & KnowledgeAA Intelligence Index 4.3.2≈5.8#302
Humanity's Last Exam3.5#328
GPQA Diamond35.1#288
FAQ
How much does Mistral Large cost?›
$2 per 1M input tokens and $6 per 1M output tokens; cached input $0.20 per 1M.
Is Mistral Large free to use?›
There is no free API; the cheapest paid price is $2 per 1M input tokens.
How good is Mistral Large, and what is it best at?›
#325 of 334 in Chat & Knowledge, #138 of 152 in Coding, #145 of 195 in Writing & Language.
Is Mistral Large better than Qwen2.5 Coder 32B Instruct?›
On Humanity's Last Exam, Mistral Large scores 3.5 vs 3.5 — about the same.
How can I use Mistral Large?›
Through the API of AWS Bedrock, Mistral.
What is the context window of Mistral Large?›
128K tokens, with up to 102K output tokens.
When was Mistral Large released? Is there a newer version?›
Released 2024-02-26. It has been superseded by Mistral Large 2407.
Is Mistral Large open source?›
No, Mistral Large is a closed model available through an API. The closest open-weights option here is MiMo v2.6 Pro.
Can Mistral Large read images?›
It accepts text and PDF files as input.
What is Mistral Large?›
Mistral Large is a closed language model from Mistral, released Feb 2024.
What are the best alternatives to Mistral Large?›
MiMo v2.6 Pro (cheaper), GPT-6 Luna (better).