Kimi K2.6 vs Kimi K3
Wins 0 of 8 areas
—
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
Kimi K3 is the stronger all-rounder.Kimi K2.6 is cheaper.
Scores updated · 41 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.
- AgentsCarrying out multi-step tasks on its own08Kimi K38 of 8 tests
- CodingWriting and fixing software05Kimi K35 of 5 tests
- ReasoningHard problems that need careful thinking04Kimi K34 of 4 tests
- MathCompetition and research-level math03Kimi K33 of 5 tests · 2 ties
- Images and chartsUnderstanding pictures, charts and video02Kimi K32 of 2 tests
- FactsGetting facts right instead of making them up12Kimi K32 of 3 tests
- Long documentsFinding answers in very long texts01Kimi K31 of 1 test
- Following instructionsDoing exactly what it is asked01Kimi K31 of 1 test
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.
Kimi K2.6 costs 73% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Kimi K3 pulls ahead
- Real work tasks from 44 professionsGDPVal+24.7points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+22.7points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+19.1points ahead
Where Kimi K2.6 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+12.7points ahead
Every test, side by side
All 41 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K3
- Terminal-Bench 2.1Kimi K3 by 19.165.985+19.1
- LiveBench · Agentic CodingKimi K3 by 15.346.962.2+15.3
- LMArena · WebDevKimi K3 by 146 rating points15091655+146 rating
- SciCodeKimi K3 by 851.559.5+8
- LiveBench · CodingKimi K3 by 2.978.681.5+2.9
AgentsKimi K3
- GDPValKimi K3 by 24.72751.7+24.7
- τ-Bench V3 · BankingKimi K3 by 22.723.346+22.7
- AA IT-Bench SREKimi K3 by 16.531.247.7+16.5
- MCP AtlasKimi K3 by 16.168.184.2+16.1
- AA ApexAgentsKimi K3 by 12.828.541.3+12.8
- Terminal-Bench 4.0Kimi K3 by 12.10.512.6+12.1
- OSWorld-VerifiedKimi K3 by 11.773.184.8+11.7
- BrowseCompKimi K3 by 4.986.391.2+4.9
ReasoningKimi K3
- CritPtKimi K3 by 15.4823.4+15.4
- LiveBench · ReasoningKimi K3 by 11.379.490.7+11.3
- Humanity's Last ExamKimi K3 by 9.437.546.9+9.4
- GPQA DiamondKimi K3 by 2.491.193.5+2.4
FactsKimi K3
- SimpleQA VerifiedKimi K3 by 15.734.950.6+15.7
- AA-Omniscience · AccuracyKimi K3 by 1532.647.6+15
- AA-Omniscience · Non-hallucinationKimi K2.6 by 12.759.546.8+12.7
Images and chartsKimi K3
- CharXiv (RQ)Kimi K3 by 4.686.791.3+4.6
- MMMU-ProKimi K3 by 1.179.480.5+1.1
MathKimi K3
- FrontierMath Tiers 1-3 (v2)Kimi K3 by 1557.272.2+15
- FrontierMath Tier 4Kimi K3 by 13.425.639+13.4
- HMMT Feb. 2026Kimi K3 by 4.392.797+4.3
- AIME 2026tie96.496.7tie
- LiveBench · Mathematicstie84.384.4tie
Following instructionsKimi K3
- LiveBench · Instruction FollowingKimi K3 by 764.471.4+7
Other results12 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.
- GDPval-AA v2Kimi K3 by 492 rating points11901682+492 rating
- AA Agentic IndexKimi K3 by 28.522.150.6+28.5
- ToolathlonKimi K3 by 23.25073.2+23.2
- BabyVisionKimi K3 by 17.268.585.7+17.2
- AA IntelligenceKimi K3 by 16.62743.6+16.6
- AA-OmniscienceKimi K3 by 14.45.319.7+14.4
- Artificial Analysis Coding IndexKimi K3 by 14.461.876.2+14.4
- livebench_data_analysisKimi K3 by 13.665.178.7+13.6
- DeepSearchQA (F1)Kimi K3 by 128395+12
- livebench_languageKimi K3 by 10.475.185.5+10.4
- HLE (with tools)Kimi K3 by 5.85459.8+5.8
- MATH-VisionKimi K3 by 4.693.297.8+4.6
Questions people ask
Which is better, Kimi K2.6 or Kimi K3?
Kimi K3 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. Kimi K2.6 wins none, but costs 73% less.
Which is better for coding?
Kimi K3. It wins 5 of the 5 coding tests both models report; Kimi K2.6 wins none.
Which is cheaper?
Kimi K2.6 costs $0.95 per million input tokens and $4.00 per million output tokens; Kimi K3 costs $3.00 and $15.00. That makes Kimi K2.6 about 73% cheaper for the same work.
How do you compare the two?
We use the 41 benchmark tests both models have published scores on. The verdict counts the 29 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 12 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.