Kimi K3 vs Mistral Large 4
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
Wins 0 of 8 areas
—
Kimi K3 is the stronger all-rounder.Mistral Large 4 is cheaper.
Scores updated · 40 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software41Kimi K34 of 5 tests
- ReasoningHard problems that need careful thinking30Kimi K33 of 3 tests
- AgentsCarrying out multi-step tasks on its own21Kimi K32 of 3 tests
- FactsGetting facts right instead of making them up21Kimi K32 of 3 tests
- MathCompetition and research-level math21Kimi K32 of 3 tests
- Images and chartsUnderstanding pictures, charts and video10Kimi K31 of 1 test
- Long documentsFinding answers in very long texts10Kimi K31 of 1 test
- Following instructionsDoing exactly what it is asked10Kimi K31 of 1 test
Images and charts, long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Mistral Large 4 costs 84% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Kimi K3 pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+28.2points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+21.8points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+13.6points ahead
Where Mistral Large 4 pulls ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+12.1points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+11.3points ahead
- Fresh competition math problemsLiveBench · Mathematics+9.2points ahead
Every test, side by side
All 40 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K3
- LMArena · WebDevKimi K3 by 88 rating points16581570+88 rating
- SciCodeKimi K3 by 5.359.554.2+5.3
- Terminal-Bench 2.1Mistral Large 4 by 5.28590.2+5.2
- LiveBench · Agentic CodingKimi K3 by 562.257.2+5
- LiveBench · CodingKimi K3 by 4.381.577.2+4.3
AgentsKimi K3
- Terminal-Bench 4.0Mistral Large 4 by 12.112.624.7+12.1
- BrowseCompKimi K3 by 6.791.284.5+6.7
- GDPValKimi K3 by 5.551.746.2+5.5
ReasoningKimi K3
- CritPtKimi K3 by 13.423.410+13.4
- Humanity's Last ExamKimi K3 by 11.946.935+11.9
- LiveBench · ReasoningKimi K3 by 6.890.783.9+6.8
FactsKimi K3
- SimpleQA VerifiedKimi K3 by 28.250.622.4+28.2
- AA-Omniscience · AccuracyKimi K3 by 21.847.625.8+21.8
- AA-Omniscience · Non-hallucinationMistral Large 4 by 11.346.858.1+11.3
MathKimi K3
- FrontierMath Tiers 1-3 (v2)Kimi K3 by 13.672.258.6+13.6
- FrontierMath Tier 4Kimi K3 by 12.23926.8+12.2
- LiveBench · MathematicsMistral Large 4 by 9.284.493.6+9.2
Following instructionsKimi K3
- LiveBench · Instruction FollowingKimi K3 by 6.671.464.8+6.6
Other results20 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- livebench_languageKimi K3 by 35.985.549.6+35.9
- Terminal-Bench 4.0Mistral Large 4 by 28.412.641+28.4
- AA-OmniscienceKimi K3 by 2519.7-5.3+25
- CyberGym-E2E-AAMistral Large 4 by 215879+21
- Terminal-Bench 4.0 (AA Coding Agent Index)Mistral Large 4 by 202141+20
- StrongREJECT (cyber subset): refusal rateMistral Large 4 by 14.877.892.6+14.8
- CritPt v2 (Hard Subset)Kimi K3 by 13.539.526+13.5
- CritPt v2 HardKimi K3 by 13.540.527+13.5
- JailbreakBench (cyber subset): safe-response rateMistral Large 4 by 10.886.597.3+10.8
- HLE-DiamondKimi K3 by 9.626.917.3+9.6
- DeepSWE 1.1Kimi K3 by 7.36961.7+7.3
- AA IntelligenceKimi K3 by 5.243.638.4+5.2
- GDP (Surge AI)Kimi K3 by 4.42217.6+4.4
- Terminal-Bench-Science 0.1Kimi K3 by 3.88.14.3+3.8
- BeyondAIMEMistral Large 4 by 2.787.990.6+2.7
- livebench_data_analysisKimi K3 by 2.278.776.5+2.2
- CyberGymMistral Large 4 by 28082+2
- BrowseComp subsetKimi K3 by 1.886.885+1.8
- BrowseComp subset (internal)Kimi K3 by 1.886.885+1.8
- AgentHarm (cyber subset): 100 minus harm scoretie100100tie
Questions people ask
Which is better, Kimi K3 or Mistral Large 4?
Kimi K3 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. Mistral Large 4 wins none, but costs 84% less.
Which is better for coding?
Kimi K3. It wins 4 of the 5 coding tests both models report; Mistral Large 4 wins 1.
Which is cheaper?
Kimi K3 costs $3.00 per million input tokens and $15.00 per million output tokens; Mistral Large 4 costs $0.70 and $2.10. That makes Mistral Large 4 about 84% cheaper for the same work.
How do you compare the two?
We use the 40 benchmark tests both models have published scores on. The verdict counts the 20 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 20 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.