87.8%
MMLU
Models / Mistral
Mistral Large 3
Mistral Large · Released 2025-12-02
Mistral's December 2025 flagship MoE. 262K context, vision input, $0.50/$1.50 per 1M tokens. 675B total / 41B active parameters.
European open-API frontier for multilingual enterprise chat, document QA, and vision+text workflows.
#152 of 168 on AA Index · current snapshot
16 AA Index · 20 benchmark rows · 6 sourcesPosition is benchmark-specific — not a cross-family or cross-source ranking.Composite48
Context262K
Input / 1M$0.50
Output / 1M$1.5
Knowledge cutoff2025-06-01
Output speed78 t/s
TTFT (default API)0.77s
Statussolid
Family profile
Best published score in each covered benchmark family.
62/100 avg
- Knowledge
- 88
- Reasoning
- 71
- Math
- 38
- Coding
- 92
- Agentic
- 12
- Long context
- 51
- Tool use
- 80
7 tested benchmark families
Benchmark placements
Per-benchmark positions
Where Mistral Large 3 places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.
Family context
Strongest rows in each family
The three highest-scoring models with pages in each capability family. Where Mistral Large 3 shows up, it's highlighted.
KnowledgeMistral Large 3 · 88
- 191
- 291
- 391
ReasoningMistral Large 3 · 71
- 195
- 295
- 395
MathMistral Large 3 · 38
- 199
- 299
- 399
CodingMistral Large 3 · 92
- 196
- 296
- 396
AgenticMistral Large 3 · 12
- 1100
- 2100
- 3100
Long contextMistral Large 3 · 51
- 183
- 283
- 381
Tool useMistral Large 3 · 80
- 189
- 287
- 387
Safety
- 192
- 292
- 391
Human preference
- 195
- 292
- 391
Newest receipts
14 recentLast 120 days
Benchmark rows added to the public ledger for Mistral Large 3 in the last 120 days. Older rows live in the full table below.
| Benchmark | Family | Score | Source | Days ago |
|---|---|---|---|---|
| AA-LCR | Long context | 34.67% | Artificial Analysis | 5 |
| AA output speed | Performance | 78 t/s | Artificial Analysis | 5 |
| AA time to first token | Performance | 0.77s | Artificial Analysis | 5 |
| AIME | Math | 38% | Artificial Analysis | 5 |
| Artificial Analysis Intelligence Index | Reasoning | 15.9% | Artificial Analysis | 5 |
| Humanity's Last Exam | Reasoning | 4.2% | Artificial Analysis | 5 |
| IFBench | Reasoning | 36.19% | Artificial Analysis | 5 |
| LiveCodeBench | Coding | 46.5% | Artificial Analysis | 5 |
| SciCode | Coding | 36.2% | Artificial Analysis | 5 |
| τ³-Bench Banking | Agentic | 5.77% | Artificial Analysis | 5 |
| Terminal-Bench | Agentic | 11.99% | Artificial Analysis | 5 |
| SWE-bench Verified | Coding | 76.2% | SWE-bench Verified leaderboard | 89 |
| GPQA Diamond | Reasoning | 71.2% | Artificial Analysis GPQA Diamond evaluation | 89 |
| MCP Atlas | Tool use | 70.4% | MCP Atlas leaderboard | 89 |
Family coverage
Best score per family
A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.
Price vs. performance
Where Mistral Large 3 sits on the scatter
Hover any dot for name, score, and input price. Keyboard: tab through the top twelve, or use the ranking below.
AnthropicOpenAIMistralAlibabaDeepSeekGoogleMoonshotMetaMiniMaxxAIZhipuXiaomiOtherMeituanCohereNVIDIA
Full benchmark ledger
Every catalog benchmark for Mistral Large 3. Scores link to the original source; gaps mean no public row exists yet.
20 of 37 catalog benchmarks have a sourced row for Mistral Large 3.
Knowledge
72.8%
MMLU-Pro
Reasoning
15.9%
AA Intelligence Index v4.1
71.2%
GPQA Diamond
Humanity's Last ExamVerified
4.2%
HLE (AA run)
IFBenchVerified
36.19%
IFBench (AA run)
62.8%
LiveBench
Math
AIMEVerified
38%
AIME 2025 (AA run)
Coding
92.4% pass@1
HumanEval
LiveCodeBenchVerified
46.5%
LiveCodeBench (AA run)
SciCodeVerified
36.2%
SciCode (AA run)
76.2%
SWE-bench Verified
Agentic
Terminal-BenchVerified
11.99%
Terminal-Bench v2.1 (AA run)
τ³-Bench BankingVerified
5.77%
τ³-Bench Banking (AA run)
Long context
AA-LCRVerified
34.67%
AA-LCR (AA run)
50.8%
LongBench v2
Tool use
Performance
AA output speedVerified
78 t/s
Median output tokens/s (1k prompt, default provider)
AA time to first tokenVerified
0.77s
Median time to first token (1k prompt, default provider)
Sources
Where the numbers come from
- Mistral model docs· checked 2026-06-09
- Mistral pricing· checked 2026-06-09