GLM-5 vs Kimi K2.5
Wins 4 of 7 areas
Coding · Agents · Facts · Following instructions
Wins 3 of 7 areas
Reasoning · Math · Long documents
GLM-5 wins more areas, narrowly.Kimi K2.5 is cheaper and better at reasoning.
Scores updated · 44 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 own40GLM-54 of 4 tests
- CodingWriting and fixing software31GLM-53 of 5 tests · 1 tie
- FactsGetting facts right instead of making them up21GLM-52 of 3 tests
- Following instructionsDoing exactly what it is asked10GLM-51 of 1 test
- ReasoningHard problems that need careful thinking04Kimi K2.54 of 4 tests
- MathCompetition and research-level math01Kimi K2.51 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts01Kimi K2.51 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 K2.5 costs 14% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GLM-5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+30.4points ahead
- Real work tasks from 44 professionsGDPVal+27.4points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+8.4points ahead
Where Kimi K2.5 pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+8.9points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+6.9points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+5.9points ahead
Every test, side by side
All 44 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGLM-5
- Terminal-Bench HardGLM-5 by 8.443.234.8+8.4
- SWE-bench ProGLM-5 by 4.455.150.7+4.4
- SciCodeKimi K2.5 by 2.846.249+2.8
- SWE-bench VerifiedGLM-5 by 177.876.8+1
- LMArena · WebDevtie14341436tie
AgentsGLM-5
- GDPValGLM-5 by 27.444.617.2+27.4
- MCP AtlasGLM-5 by 467.863.8+4
- AA ApexAgentsGLM-5 by 314.511.5+3
- BrowseCompGLM-5 by 175.974.9+1
ReasoningKimi K2.5
- ARC-AGI-2Kimi K2.5 by 6.94.911.8+6.9
- GPQA DiamondKimi K2.5 by 5.98287.9+5.9
- Humanity's Last ExamKimi K2.5 by 1.429.330.7+1.4
- CritPtKimi K2.5 by 1.123.1+1.1
FactsGLM-5
- AA-Omniscience · Non-hallucinationGLM-5 by 30.464.734.3+30.4
- AA-Omniscience · AccuracyKimi K2.5 by 8.926.335.2+8.9
- Vectara HHEM hallucination ratelower is betterGLM-5 by 4.110.114.2+4.1
MathKimi K2.5
- HMMT Feb. 2026Kimi K2.5 by 4.382.887.1+4.3
- IMOAnswerBenchtie82.581.8tie
- AIME 2026tie95.495.8tie
Other results23 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.
- Vending-Bench 2GLM-5 by 3234 rating points44321198+3234 rating
- AA Agentic IndexGLM-5 by 41.463.121.7+41.4
- ARC-AGI-1Kimi K2.5 by 20.644.765.3+20.6
- Tool-DecathlonGLM-5 by 10.23827.8+10.2
- HMMT Nov. 2025GLM-5 by 7.796.989.2+7.7
- AA-OmniscienceGLM-5 by 7.60.3-7.3+7.6
- vectara_answer_rateGLM-5 by 7.599.792.2+7.5
- CyberGymGLM-5 by 748.341.3+7
- Terminal-Bench 2.0GLM-5 by 5.456.250.8+5.4
- AA IntelligenceGLM-5 by 4.427.923.5+4.4
- τ²-BenchGLM-5 by 4.389.785.4+4.3
- vectara_factual_consistencyGLM-5 by 4.189.985.8+4.1
- NL2RepoGLM-5 by 3.935.932+3.9
- vectara_avg_summary_lengthKimi K2.5 by 37.6 rating points74.4112+37.6 rating
- τ³-BenchGLM-5 by 3.269.266+3.2
- Artificial Analysis Coding IndexKimi K2.5 by 2.644.246.8+2.6
- τ²-Bench Telecom (AA run)GLM-5 by 2.398.295.9+2.3
- frontiermath_tier_4_v1Kimi K2.5 by 2.12.14.2+2.1
- swe_bench_bashGLM-5 by 272.870.8+2
- browsecomp_with_context_managerGLM-5 by 175.974.9+1
- AIME 2025tie96.796.1tie
- AIME 2026 I (Tools-allowed)tie92.792.5tie
- HLE (with tools)tie50.450.2tie
Questions people ask
Which is better, GLM-5 or Kimi K2.5?
GLM-5 wins four of the seven areas where both have results: coding, agents, facts and following instructions. Kimi K2.5 wins reasoning, math and long documents, and costs 14% less.
Which is better for coding?
GLM-5. It wins 3 of the 5 coding tests both models report; Kimi K2.5 wins 1, and 1 is a tie.
Which is cheaper?
GLM-5 costs $1.00 per million input tokens and $3.20 per million output tokens; Kimi K2.5 costs $0.60 and $3.00. That makes Kimi K2.5 about 14% cheaper for the same work.
How do you compare the two?
We use the 44 benchmark tests both models have published scores on. The verdict counts the 21 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 23 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.