Mistral Large 3 675B Instruct 2512 — Benchmarks

Benchmark scores for Mistral Large 3 675B Instruct 2512 aggregated from public leaderboards (7 boards), with how it ranks among open models and how it compares with its nearest neighbours. See hardware requirements for what you need to run it.

llmrun Composite Score →

89.6

90% range 81.5–97.6 · from 5 boards

Coding92.890% range 76–110
AgentsNot enough data yet
MathNot enough data yet
Science96.990% range 75–119
Reasoning89.090% range 80–98

Mistral Large 3 675B Instruct 2512 reasoning benchmarks

Multi-step logic and problem-solving — can the model think through a hard problem rather than recall an answer.

CritPt
CritPt (Critical Points) tests graduate-level physics problem-solving on research-style questions spanning mechanics, electromagnetism, quantum mechanics, and more. It was created by physicists to probe whether a model can reason through multi-step physics rather than pattern-match textbook answers.
0.0%
0.0%
median 0.9%open leader 26.6%
Overall best: Gemini 4 Carbon (high) 36.0%
19th percentile · #72 of 84 open · #227 of 244 all
#72 of 84
#227 of 244
Lab-reported
Epoch AI · Dec 2025
DTBench
DTBench is a set of handcrafted, expert-validated multiple-choice questions on decision theory — Newcomb-like problems and related puzzles — built by the Conceptual Reasoning Index project with Anthropic. It tests whether a model can follow an abstract argument rather than recall a fact.
65.1%
65.1%
median 68.7%open leader 93.9%
Overall best: Claude Opus 5.5 Max 98.9%
45th percentile · #39 of 70 open · #172 of 235 all
#39 of 70
#172 of 235
Lab-reported
Epoch AI · Dec 2025
LMCA
LMCA (Language Model Conceptual Argumentation) scores models on expert-rated arguments across philosophy, decision theory, and AI risk — domains where there is no empirical feedback loop to check an answer against. It is part of the Conceptual Reasoning Index, built with Anthropic by researchers Emery Cooper and Caspar Oesterheld.
16.7%
16.7%
median 25.5%open leader 55.5%
Overall best: Claude Sonnet 5.5 Max 68.3%
29th percentile · #40 of 56 open · #171 of 197 all
#40 of 56
#171 of 197
Lab-reported
Epoch AI · Dec 2025

Mistral Large 3 675B Instruct 2512 coding benchmarks

Writing, editing, and fixing real code — measured by whether the resulting program actually runs and passes tests.

SciCode
SciCode asks a model to write real scientific-computing code — implementing multi-step numerical methods drawn from physics, chemistry, biology, and math research — and checks the output against reference solutions. It was built by a team of scientists and ML researchers to test coding ability on genuine research problems, not typical software tasks.
36.6%
36.6%
median 40.1%open leader 60.9%
Overall best: Claude Opus 5.5 Max 66.9%
35th percentile · #47 of 72 open · #187 of 217 all
#47 of 72
#187 of 217
Lab-reported
Epoch AI · Dec 2025
ALE-Bench
ALE-Bench, from Sakana AI and AtCoder, tests long-horizon algorithmic optimisation using problems drawn from AtCoder Heuristic Contests, where there is no single correct answer and solutions are scored on quality. Results are reported as an AtCoder-style performance rating, where higher is better.
264.7
264.7
median 754.6open leader 1932.8
Overall best: GPT 6 Astra Max 2951.3
4th percentile · #46 of 48 open · #123 of 129 all
#46 of 48
#123 of 129
Lab-reported
Epoch AI · Dec 2025

Other results (1)leaderboard-only, not part of the llmrun Score

Leaderboard-only, not part of the llmrun Score

LMArena WebDev
Code Arena (formerly WebDev Arena) is Arena's head-to-head coding leaderboard: two models each build a real web app or UI from the same prompt, and people vote blind on which result is better. The rating shown is a relative Bradley-Terry score, not a percentage — compare it only to other models on this board.
1232.6
1232.6
median 1415.2open leader 1665.0
Overall best: Claude Opus 5.5 (Max) 1809.8
4th percentile · #53 of 55 open · #138 of 146 all
#53 of 55
#138 of 146
Arena
Arena · Oct 2026

Mistral Large 3 675B Instruct 2512 instruction following benchmarks

How precisely a model obeys explicit, verifiable instructions — format, length, and content constraints — and, more broadly, how much people prefer its responses head-to-head in blind human voting.

LMArena Text
Arena (arena.ai, formerly LMArena / Chatbot Arena) ranks models by millions of blind, head-to-head human votes on real prompts. The Arena Score shown here is a Bradley-Terry rating computed with Arena's default style control, so higher means people preferred that model's answers more often — it's a relative rating, not a percentage, so compare it only to other models on this board.
1414.3
1414.3
median 1321.4open leader 1488.2
Overall best: Gemini 4 Carbon (High) 1527.5
79th percentile · #41 of 188 open · #139 of 408 all
#41 of 188
#139 of 408
Arena
Arena · Oct 2026

Score vs VRAM in the ≤512 GB tier

Mistral Large 3 675B Instruct 2512
Other models
Best Score for its VRAM
x: VRAM (GB, log scale) · y: llmrun Score

#71 of 102 that fit ≤512 GB

Each dot is a model at the best quant that fits ≤512 GB. The line steps up to the best Score available at each amount of VRAM; a dot below it is beaten by a model that needs no more memory.

Full ≤512 GB ranking →

Mistral Large 3 675B Instruct 2512 vs its neighbours

Same boards for the nearest open models per llmrun Score. The highest score in each row is marked.

Board	Mistral Large 3 675B Instruct 2512this model	Qwen3.8 27Bbest in classQ4_K_M · 18 GB	DeepSeek v3 0324nearest by ScoreQ4_K_M · 412 GB	DeepSeek v3nearest by ScoreQ4_K_M · 412 GB	Llama 4 Maverick 17B 128E Instructnearest by ScoreQ4_K_M · 273 GB
llmrun Score	89.6	122.0	93.9	89.0	87.1
CritPt	0.0%	5.4%	0.0%	0.0%	0.0%
DTBench	65.1%	88.0%	64.8%	—	61.9%
LMCA	16.7%	41.4%	15.5%	—	15.9%
SciCode	36.6%	46.6%	39.0%	35.8%	31.7%
ALE-Bench	264.7	—	—	—	173.0
LMArena Text	1414.3	1435.6	1395.2	1358.4	1327.0
Compare head to head		Compare	Compare	Compare	Compare

Scores aggregated from public benchmark sources (Epoch AI: 5 boards, Arena: 2 boards) — each linked from the benchmark pages. llmrun does not run these benchmarks. Percentiles and ranks are among open models in the llmrun catalogue; the median and leader marks refer to the same field.