GLM-5.2 vs Kimi K3
Wins 0 of 7 areas
—
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
Kimi K3 is the stronger all-rounder.GLM-5.2 is cheaper.
Scores updated · 59 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. 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 own06Kimi K36 of 6 tests
- ReasoningHard problems that need careful thinking06Kimi K36 of 6 tests
- CodingWriting and fixing software05Kimi K35 of 5 tests
- MathCompetition and research-level math23Kimi K33 of 5 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.
GLM-5.2 costs 68% 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
- Abstract visual puzzles that people can solveARC-AGI-2+37.6points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+23.3points ahead
- Short factual questions, answered correctlySimpleQA Verified+16.4points ahead
Where GLM-5.2 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+26.9points ahead
- Fresh competition math problemsLiveBench · Mathematics+5.4points ahead
- US invitational high-school math exam problemsAIME 2026+2.5points ahead
Every test, side by side
All 59 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K3
- LiveBench · Agentic CodingKimi K3 by 10.451.862.2+10.4
- SciCodeKimi K3 by 8.351.259.5+8.3
- Terminal-Bench 2.1Kimi K3 by 7.177.985+7.1
- LMArena · WebDevKimi K3 by 52 rating points16031655+52 rating
- LiveBench · CodingKimi K3 by 1.879.781.5+1.8
AgentsKimi K3
- Terminal-Bench 4.0Kimi K3 by 11.6112.6+11.6
- τ-Bench V3 · BankingKimi K3 by 11.434.646+11.4
- GDPValKimi K3 by 843.751.7+8
- AA ApexAgentsKimi K3 by 7.633.741.3+7.6
- MCP AtlasKimi K3 by 7.476.884.2+7.4
- AA IT-Bench SREKimi K3 by 542.747.7+5
ReasoningKimi K3
- ARC-AGI-2Kimi K3 by 37.622.860.4+37.6
- LiveBench · ReasoningKimi K3 by 12.178.690.7+12.1
- Humanity's Last ExamKimi K3 by 5.841.146.9+5.8
- GPQA DiamondKimi K3 by 489.593.5+4
- CritPtKimi K3 by 2.520.923.4+2.5
- SimpleBenchKimi K3 by 1.958.860.7+1.9
FactsKimi K3
- AA-Omniscience · Non-hallucinationGLM-5.2 by 26.973.746.8+26.9
- AA-Omniscience · AccuracyKimi K3 by 23.324.347.6+23.3
- SimpleQA VerifiedKimi K3 by 16.434.250.6+16.4
MathKimi K3
- FrontierMath Tiers 1-3 (v2)Kimi K3 by 1359.272.2+13
- FrontierMath Tier 4Kimi K3 by 9.729.339+9.7
- LiveBench · MathematicsGLM-5.2 by 5.489.884.4+5.4
- HMMT Feb. 2026Kimi K3 by 4.592.597+4.5
- AIME 2026GLM-5.2 by 2.599.296.7+2.5
Following instructionsKimi K3
- LiveBench · Instruction FollowingKimi K3 by 9.162.371.4+9.1
Other results32 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.
- SWE-MarathonKimi K3 by 291342+29
- DeepSWE 1.1Kimi K3 by 254469+25
- ToolathlonKimi K3 by 2548.273.2+25
- OfficeQA ProKimi K3 by 21.941.463.3+21.9
- DeepSWEKimi K3 by 21.346.267.5+21.3
- AutomationBench PublicKimi K3 by 17.912.930.8+17.9
- GDPval-AA v2 EloKimi K3 by 176 rating points15101686+176 rating
- ARC-AGI-1Kimi K3 by 17.57794.5+17.5
- GDPval-AA v2Kimi K3 by 174 rating points15081682+174 rating
- Toolathlon VerifiedKimi K3 by 16.659.976.5+16.6
- AA-OmniscienceKimi K3 by 15.34.419.7+15.3
- Program BenchKimi K3 by 14.163.777.8+14.1
- Terminal-bench 3.0Kimi K3 by 12.84.617.4+12.8
- DSBench-FullStackKimi K3 by 11.961.873.7+11.9
- AA Agentic IndexKimi K3 by 11.239.450.6+11.2
- AA IntelligenceKimi K3 by 9.933.743.6+9.9
- JobBenchKimi K3 by 9.543.452.9+9.5
- livebench_languageKimi K3 by 9.376.285.5+9.3
- NL2RepoKimi K3 by 9.148.958+9.1
- DSBench-HardKimi K3 by 8.554.563+8.5
- ProgramBench (Almost Solved)Kimi K3 by 89.517.5+8
- MLS Bench Litelower is betterGLM-5.2 by 7.940.448.3+7.9
- Artificial Analysis Coding IndexKimi K3 by 7.468.876.2+7.4
- SpreadsheetBench 2Kimi K3 by 6.728.134.8+6.7
- HLE (with tools)Kimi K3 by 5.154.759.8+5.1
- ResearchRubricsKimi K3 by 5.171.176.2+5.1
- livebench_data_analysisKimi K3 by 573.778.7+5
- Agents' Last ExamKimi K3 by 4.523.828.3+4.5
- Agents' Last Exam (ALE-CLI)Kimi K3 by 3.823.827.6+3.8
- CyberGymKimi K3 by 2.877.280+2.8
- PostTrainBenchKimi K3 by 2.334.336.6+2.3
- HLE (wo / w tools)Kimi K3 by 1.354.756+1.3
Questions people ask
Which is better, GLM-5.2 or Kimi K3?
Kimi K3 wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. GLM-5.2 wins none, but costs 68% less.
Which is better for coding?
Kimi K3. It wins 5 of the 5 coding tests both models report; GLM-5.2 wins none.
Which is cheaper?
GLM-5.2 costs $1.40 per million input tokens and $4.40 per million output tokens; Kimi K3 costs $3.00 and $15.00. That makes GLM-5.2 about 68% cheaper for the same work.
How do you compare the two?
We use the 59 benchmark tests both models have published scores on. The verdict counts the 27 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 32 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.