GPT-5.4 vs Kimi K2.5
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
Wins 0 of 8 areas
—
GPT-5.4 is the stronger all-rounder.Kimi K2.5 is cheaper.
Scores updated · 62 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 own60GPT-5.46 of 6 tests
- CodingWriting and fixing software51GPT-5.45 of 6 tests
- ReasoningHard problems that need careful thinking40GPT-5.44 of 4 tests
- Images and chartsUnderstanding pictures, charts and video30GPT-5.43 of 3 tests
- MathCompetition and research-level math30GPT-5.43 of 3 tests
- FactsGetting facts right instead of making them up31GPT-5.43 of 4 tests
- Long documentsFinding answers in very long texts10GPT-5.41 of 1 test
- Following instructionsDoing exactly what it is asked10GPT-5.41 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 79% 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 GPT-5.4 pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+61.5points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+32.6points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+25.4points ahead
Where Kimi K2.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+26points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+1.3points ahead
Every test, side by side
All 62 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.4
- Terminal-Bench 2.1GPT-5.4 by 32.678.345.7+32.6
- Terminal-Bench HardGPT-5.4 by 22.857.634.8+22.8
- SciCodeGPT-5.4 by 7.656.649+7.6
- SWE-bench ProGPT-5.4 by 757.750.7+7
- LMArena · WebDevGPT-5.4 by 29 rating points14651436+29 rating
- SWE-bench MultilingualKimi K2.5 by 1.371.773+1.3
AgentsGPT-5.4
- τ-Bench V3 · BankingGPT-5.4 by 25.439.614.2+25.4
- AA ApexAgentsGPT-5.4 by 21.833.311.5+21.8
- GDPValGPT-5.4 by 20.237.417.2+20.2
- OSWorld-VerifiedGPT-5.4 by 11.77563.3+11.7
- BrowseCompGPT-5.4 by 7.882.774.9+7.8
- MCP AtlasGPT-5.4 by 3.467.263.8+3.4
ReasoningGPT-5.4
- ARC-AGI-2GPT-5.4 by 61.573.311.8+61.5
- CritPtGPT-5.4 by 20.323.43.1+20.3
- Humanity's Last ExamGPT-5.4 by 1343.730.7+13
- GPQA DiamondGPT-5.4 by 4.19287.9+4.1
FactsGPT-5.4
- AA-Omniscience · Non-hallucinationKimi K2.5 by 268.334.3+26
- AA-Omniscience · AccuracyGPT-5.4 by 15.750.935.2+15.7
- SimpleQA VerifiedGPT-5.4 by 10.845.134.3+10.8
- Vectara HHEM hallucination ratelower is betterGPT-5.4 by 7.2714.2+7.2
Images and chartsGPT-5.4
- CharXiv (RQ)GPT-5.4 by 5.382.877.5+5.3
- LMArena · VisionGPT-5.4 by 33 rating points13021269+33 rating
- MMMU-ProGPT-5.4 by 378.475.4+3
MathGPT-5.4
- HMMT Feb. 2026GPT-5.4 by 10.697.787.1+10.6
- IMOAnswerBenchGPT-5.4 by 9.691.481.8+9.6
- AIME 2026GPT-5.4 by 3.499.295.8+3.4
Other results34 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.
- baby_vision_with_pythonGPT-5.4 by 39.780.240.5+39.7
- MCPMarkGPT-5.4 by 3362.529.5+33
- ARC-AGI-1GPT-5.4 by 28.493.765.3+28.4
- Tool-DecathlonGPT-5.4 by 26.854.627.8+26.8
- ToolathlonGPT-5.4 by 26.854.627.8+26.8
- CyberGymGPT-5.4 by 2566.341.3+25
- Artificial Analysis Coding IndexGPT-5.4 by 24.371.146.8+24.3
- Terminal-Bench 2.0GPT-5.4 by 24.375.150.8+24.3
- frontiermath_tier_4_v1GPT-5.4 by 22.927.14.2+22.9
- AA Agentic IndexGPT-5.4 by 22.544.221.7+22.5
- AA IntelligenceGPT-5.4 by 15.53923.5+15.5
- DeepSearchQA (accuracy)Kimi K2.5 by 13.463.777.1+13.4
- BabyVisionGPT-5.4 by 13.249.736.5+13.2
- AA-OmniscienceGPT-5.4 by 13.15.8-7.3+13.1
- matharena_visual_math_overallGPT-5.4 by 11.992.580.6+11.9
- V* (w/ python)GPT-5.4 by 11.598.486.9+11.5
- charxiv_rq_with_pythonGPT-5.4 by 11.39078.7+11.3
- LiveBenchGPT-5.4 by 11.280.369.1+11.2
- mathvision_with_pythonGPT-5.4 by 11.196.185+11.1
- NL2RepoGPT-5.4 by 9.341.332+9.3
- τ²-Bench Telecom (AA run)Kimi K2.5 by 8.887.195.9+8.8
- browsecomp_with_context_managerGPT-5.4 by 7.882.774.9+7.8
- MATH-VisionGPT-5.4 by 7.89284.2+7.8
- vectara_answer_rateGPT-5.4 by 7.799.992.2+7.7
- vectara_factual_consistencyGPT-5.4 by 7.29385.8+7.2
- τ³-BenchGPT-5.4 by 6.972.966+6.9
- HMMT Nov. 2025GPT-5.4 by 6.695.889.2+6.6
- mmmu_pro_with_pythonGPT-5.4 by 4.482.177.7+4.4
- BrowseComp (Agent Swarm)GPT-5.4 by 4.382.778.4+4.3
- vectara_avg_summary_lengthKimi K2.5 by 30.3 rating points81.7112+30.3 rating
- HLE (with tools)GPT-5.4 by 1.85250.2+1.8
- DeepSearchQA (F1)GPT-5.4 by 1.578.677.1+1.5
- MMLU-Protie87.587.1tie
- OmniDocBench 1.5tie89.188.8tie
Questions people ask
Which is better, GPT-5.4 or Kimi K2.5?
GPT-5.4 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. Kimi K2.5 wins none, but costs 79% less.
Which is better for coding?
GPT-5.4. It wins 5 of the 6 coding tests both models report; Kimi K2.5 wins 1.
Which is cheaper?
GPT-5.4 costs $2.50 per million input tokens and $15.00 per million output tokens; Kimi K2.5 costs $0.60 and $3.00. That makes Kimi K2.5 about 79% cheaper for the same work.
How do you compare the two?
We use the 62 benchmark tests both models have published scores on. The verdict counts the 28 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 34 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.